CN117911057A - Second-hand car price prediction method, device, equipment and storage medium - Google Patents

Abstract

The invention discloses a method, a device, equipment and a storage medium for predicting the price of a second-hand vehicle, wherein the method comprises the following steps: inputting target second-hand vehicle data into a target second-hand vehicle price prediction model to obtain a second-hand vehicle price prediction value, wherein the target second-hand vehicle price prediction model is obtained by introducing Chebyshev chaotic mapping, nonlinear decreasing disturbance factors and self-adaptive weight factors to improve a Harris eagle algorithm, and optimizing super-parameters of a LightGBM model based on the improved Harris eagle algorithm. According to the invention, the Chebyshev chaotic map, the nonlinear decreasing disturbance factor and the self-adaptive weight factor are introduced to improve the Harris eagle algorithm, and then the secondary handcart price prediction is carried out based on the target secondary handcart price prediction model obtained by optimizing the super parameters of LightGBM by utilizing the improved Harris eagle algorithm, so that the accuracy and stability of secondary handcart price prediction are effectively improved.

Description

Second-hand car price prediction method, device, equipment and storage medium

Technical Field

The present invention relates to the field of data processing technology, and in particular to a second-hand car price prediction method, device, equipment and storage medium.

Background technique

With the rapid development of my country's used car market, as of 2021, the growth rate of used car sales in my country has surpassed that of new car sales, reaching 22.62%. However, as the used car market develops rapidly, the rationality of used car price evaluation has become increasingly apparent. Traditional manual evaluation methods are based on industry experience, so there are shortcomings such as high evaluation costs and arbitrary evaluation results. At the same time, information asymmetry between consumers and car dealers also makes it impossible for consumers to determine the price of used cars. Therefore, it is particularly important to establish a reasonable used car price prediction model driven by historical data to regulate the used car market. It can provide a reference for negotiations between consumers and car dealers, and can also promote the steady development of the used car market.

At present, the models that use machine learning algorithms to predict used car prices mainly include: KNN, random forest, SVR and linear regression models, etc., which can provide effective evaluation of used car price prediction. However, there is still room for improvement in its prediction accuracy. The reason is that the prediction performance of traditional machine learning models is affected by the selection of hyperparameters, and traditional manual parameter setting or grid method and gradient descent method cannot make the performance of the prediction model reach the optimal level.

Therefore, there is an urgent need for a used car price prediction method that can effectively improve the accuracy and stability of used car price prediction.

Summary of the invention

The main purpose of the present invention is to provide a second-hand car price prediction method, device, equipment and storage medium, aiming to solve the technical problem of low accuracy and stability of second-hand car price prediction in the prior art.

To achieve the above object, the present invention provides a method for predicting the price of a used car, the method comprising the following steps:

Acquire vehicle attribute information of a used car, and pre-process the vehicle attribute information to obtain target used car data;

The target used car data is input into a target used car price prediction model to obtain a used car price prediction value. The target used car price prediction model is a model obtained by optimizing the hyperparameters of the LightGBM model based on the improved Harris Eagle algorithm after introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to improve the Harris Eagle algorithm.

Optionally, before the step of acquiring the vehicle attribute information of the used car and preprocessing the vehicle attribute information to obtain the target used car data, the step further includes:

Obtaining a used car sales data set, and preprocessing the used car sales data set to obtain a training data set;

The improved Harris Hawk algorithm is obtained by introducing Chebyshev chaotic mapping, nonlinear decreasing disturbance factor and adaptive weight factor to improve the Harris Hawk algorithm.

The improved Harris Eagle algorithm is used to optimize the hyperparameters of the LightGBM model to obtain the target parameter combination;

After the initial LightGBM model is established based on the target parameter combination, the initial LightGBM model is trained using the training data set to obtain a target used car price prediction model.

Optionally, the step of improving the Harris Hawk algorithm by introducing a Chebyshev chaotic map, a nonlinear decreasing disturbance factor and an adaptive weight factor to obtain an improved Harris Hawk algorithm comprises:

Introducing Chebyshev chaotic mapping into the global exploration phase of the Harris Eagle algorithm to obtain an improved global exploration phase;

The prey escape energy in the Harris Hawk algorithm is improved by using a nonlinear decreasing disturbance factor to obtain an improved escape energy formula;

The optimal individual in the population of the Harris Hawk algorithm is weighted by using an adaptive weight factor to obtain an optimal individual update formula;

An improved Harris Eagle algorithm is obtained based on the improved global exploration phase, the improved escape energy formula and the optimal individual update formula.

Optionally, the Chebyshev chaotic map is:

CM＝cos(tcos ^-1 (γ))

In the formula, t is the current iteration number, γ is the initial value of chaos;

The improved model of the global exploration stage is:

Where X(t+1) and X(t) are the next iteration position and current position of Harris's hawk, respectively; X _rabbit (t) is the location of the prey; CM ₁ , CM ₂ , CM ₃ , CM ₄ , and q are all random numbers between (0,1); UB and LB are the upper and lower limits of the individual position search, respectively; X _m (t) is the average position of the population;

Accordingly: the improved escape energy formula is:

E＝2E ₀ (2r(0.5+cos((π(t/T)) ^1/2 )));

Where, E ₀ is the initial escape energy, T is the maximum number of iterations, t is the current number of iterations, and r is a random number between (0,1);

The adaptive weight factor is:

The optimal individual update formula:

X′ _rabbit =w(t)×X _rabbit .

Optionally, the step of obtaining a used car sales data set and preprocessing the used car sales data set to obtain a training data set includes:

Acquire a used car sales data set, preprocess the used car sales data set, and obtain a preprocessed used car sales data set;

Splitting the preprocessed used car sales data set according to a preset ratio to obtain a training data set and a test data set;

Accordingly, after the step of training the initial LightGBM model using the training data set to obtain the target used car price prediction model, the method further includes:

Using the target used car price prediction model to predict the test data set to obtain a used car price prediction value;

An evaluation index is selected to evaluate the target used car price prediction model based on the used car price prediction value, wherein the evaluation index includes a determination coefficient, a mean square error, and a mean absolute error.

Optionally, the step of optimizing the hyperparameters of the LightGBM model using the improved Harris Eagle algorithm to obtain a target parameter combination includes:

Setting the population size, number of iterations and problem dimension of the improved Harris Hawk algorithm to obtain a target Harris Hawk algorithm;

The number of leaf nodes per tree, the maximum depth of each tree, the model learning rate, and the minimum number of samples required for each leaf node in the LightGBM model are used as hyperparameters of the LightGBM model;

Input the training data set into the LightGBM model, and determine the individual fitness value according to a preset fitness function;

Based on the individual fitness values, the population individuals are iteratively updated according to the target Harris Hawk algorithm until the number of iterations is reached to obtain a target parameter combination.

Optionally, after establishing the initial LightGBM model based on the target parameter combination, the step of training the initial LightGBM model using the training data set to obtain a target used car price prediction model includes:

Establishing an initial LightGBM model based on the LightGBM model and the target parameter combination;

Using the training data set to train the initial LightGBM model to obtain a training result;

The initial LightGBM model is optimized according to the training results to obtain a target used car price prediction model.

In addition, to achieve the above-mentioned purpose, the present invention also proposes a second-hand car price prediction device, the device comprising:

A data processing module, used to obtain the vehicle attribute information of a used car, and pre-process the vehicle attribute information to obtain target used car data;

The price output module is used to input the target used car data into the target used car price prediction model to obtain the used car price prediction value. The target used car price prediction model is a model obtained by improving the Harris Eagle algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor, and optimizing the hyperparameters of the LightGBM model based on the improved Harris Eagle algorithm.

In addition, to achieve the above-mentioned purpose, the present invention also proposes a used car price prediction device, which includes: a memory, a processor, and a used car price prediction program stored in the memory and executable on the processor, wherein the used car price prediction program is configured to implement the steps of the used car price prediction method described above.

In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, on which a used car price prediction program is stored. When the used car price prediction program is executed by a processor, the steps of the used car price prediction method described above are implemented.

The present invention obtains target second-hand car data by acquiring vehicle attribute information of used cars and preprocessing the vehicle attribute information; the target second-hand car data is input into a target second-hand car price prediction model to obtain a second-hand car price prediction value, wherein the target second-hand car price prediction model is a model obtained by optimizing the hyperparameters of the LightGBM model based on the improved Harris Hawk algorithm after introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to improve the Harris Hawk algorithm. Compared with the prior art, the present invention improves the global exploration and local development capabilities of the Harris Hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor, thereby improving the global search accuracy of the algorithm, and then optimizing the hyperparameters of LightGBM based on the improved Harris Hawk algorithm, thereby solving the hyperparameter dependency problem in model prediction, thereby effectively improving the accuracy and stability of second-hand car price prediction.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of the structure of a second-hand car price prediction device in a hardware operating environment according to an embodiment of the present invention;

FIG2 is a flow chart of a first embodiment of a method for predicting used car prices according to the present invention;

FIG3 is a flow chart of a second embodiment of a method for predicting used car prices according to the present invention;

FIG4 is a schematic diagram of the improved Harris Eagle algorithm flow in the second-hand car price prediction method of the present invention;

FIG5 is a schematic diagram of a flow chart of a third embodiment of a method for predicting used car prices according to the present invention;

FIG6 is a schematic diagram of the overall design of a target second-hand car price prediction model in the second-hand car price prediction method of the present invention;

FIG. 7 is a structural block diagram of the first embodiment of the second-hand car price prediction device of the present invention.

The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

Refer to FIG. 1 , which is a schematic diagram of the structure of a used car price prediction device in a hardware operating environment according to an embodiment of the present invention.

As shown in FIG1 , the second-hand car price prediction device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk storage. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

Those skilled in the art will appreciate that the structure shown in FIG. 1 does not constitute a limitation on the used car price prediction device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

As shown in FIG. 1 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a used car price prediction program.

In the used car price prediction device shown in Figure 1, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the used car price prediction device of the present invention can be set in the used car price prediction device, and the used car price prediction device calls the used car price prediction program stored in the memory 1005 through the processor 1001, and executes the used car price prediction method provided by the embodiment of the present invention.

An embodiment of the present invention provides a method for predicting the price of a used car. Referring to FIG. 2 , FIG. 2 is a flow chart of a first embodiment of the method for predicting the price of a used car of the present invention.

In this embodiment, the second-hand car price prediction method includes the following steps:

Step S10: Acquire vehicle attribute information of a used car, and pre-process the vehicle attribute information to obtain target used car data.

It should be noted that the execution subject of this embodiment can be a computer server device with data processing, network communication and program running functions, such as a server, a tablet computer, a personal computer, an iPad, etc., or an electronic device capable of realizing the above functions, a second-hand car price prediction device, etc. The second-hand car price prediction device is taken as an example to illustrate this embodiment and the following embodiments.

It should be understood that the vehicle attribute information of the used car includes but is not limited to the car model, registration date, gearbox type, mileage, fuel type, road tax, fuel consumption, displacement and selling price.

It is understandable that the above-mentioned pre-processing of the vehicle attribute information may be data cleaning or data filling of the missing data of the vehicle attribute information of used cars to obtain the target used car data.

Step S20: Input the target used car data into the target used car price prediction model to obtain a used car price prediction value. The target used car price prediction model is a model obtained by improving the Harris Eagle algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor, and optimizing the hyperparameters of the LightGBM model based on the improved Harris Eagle algorithm.

It should be explained that the Harris Hawk Algorithm (HHO) is a metaheuristic algorithm proposed by Heidari et al. in 2019 to simulate the collaborative predation behavior of Harris Hawks. The algorithm has the advantages of few parameters, easy adjustment and implementation, and has excellent performance in many optimization problems. The HHO algorithm is divided into three stages: global exploration, transition from exploration to development, and local development.

It should be noted that chaotic sequences have the characteristics of ergodicity and randomness, which can make the distribution of populations in space more uniform, and help improve the search ability of the algorithm. In the exploration phase of the traditional Harris Hawk algorithm, the Harris Hawk observes prey randomly according to the traditional formula, resulting in the Harris Hawk population being not evenly distributed in space. Therefore, this embodiment introduces Chebyshev chaotic mapping to improve the position of the Harris Hawk in the exploration phase, so that the distribution of the Harris Hawk population in space is more uniform.

The escape energy E plays a key role in balancing the exploration and development behaviors of the Harris Hawk. In the traditional Harris Hawk algorithm, the linearly decreasing escape energy E easily causes the Harris Hawk algorithm to fall into a local optimum during the late development of iterations. Therefore, this embodiment uses a nonlinear decreasing disturbance factor to improve the prey escape energy, so that the Harris Hawk algorithm still has a certain exploration behavior in the later stage, and enhances the ability of the Harris Hawk algorithm to jump out of the local optimum.

The inertia weight has the function of balancing global search and local development. A larger inertia weight can expand the global search range of the algorithm and improve the global exploration capability. A smaller inertia weight can enable the algorithm to perform a more detailed search near the optimal solution, thereby improving the local development capability of the algorithm. Therefore, in order to enhance the ability of the Harris Hawk to approach the optimal solution within a limited range, this embodiment proposes a new adaptive weight factor to weight the optimal individuals in the population.

After the above-mentioned improvements to the Harris Hawk algorithm (HHO algorithm), the improved Harris Hawk algorithm (iHHO algorithm) is used to optimize the hyperparameters in the LightGBM model, and the target used car price prediction model is established based on this.

It should be noted that the Lightweight Gradient Boosting Machine LightGBM is an improved algorithm of the Gradient Boosting Decision Tree GBDT. The core idea of the GBDT algorithm is that each input depends on the results of the previous training. It needs to iterate the entire data set many times during the entire training process, which consumes huge time and memory. The traditional XGBoost uses the pre-sorting idea to find the best split point during training, but when faced with massive data and high-dimensional data, its time and space overhead is still large. The LightGBM model effectively solves the shortcomings of GBDT and XGBoost when faced with high dimensions and massive data, and has the advantages of fast training speed, low memory consumption and high accuracy.

This embodiment obtains target second-hand car data by acquiring vehicle attribute information of used cars and preprocessing the vehicle attribute information; the target second-hand car data is input into a target second-hand car price prediction model to obtain a second-hand car price prediction value, wherein the target second-hand car price prediction model is a model obtained by optimizing the hyperparameters of the LightGBM model based on the improved Harris Hawk algorithm after introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to improve the Harris Hawk algorithm's global exploration and local development capabilities, thereby improving the algorithm's global search accuracy, and then optimizing the hyperparameters of LightGBM based on the improved Harris Hawk algorithm, thereby solving the hyperparameter dependency problem in model prediction, thereby effectively improving the accuracy and stability of second-hand car price prediction.

Refer to FIG3 , which is a flow chart of a second embodiment of a method for predicting used car prices according to the present invention.

Based on the first embodiment above, in this embodiment, before step S10, the following steps are further included:

Step S01: Obtain a used car sales data set, and preprocess the used car sales data set to obtain a training data set.

It is understandable that the above-mentioned second-hand car sales dataset can be a second-hand car sales dataset obtained through a data open platform or web crawler technology, and this embodiment does not limit this. This embodiment and the following embodiments are described by taking the VW second-hand car sales dataset in the kaggle data open platform as an example.

It should be noted that the above data preprocessing may be to perform unique-hot encoding processing on the vehicle model, variable box type, and fuel type in the used car sales data set, convert the registration date into the corresponding years, or perform data cleaning, which is not limited in this embodiment.

Step S02: The Harris Hawk algorithm is improved by introducing Chebyshev chaotic mapping, nonlinear decreasing disturbance factor and adaptive weight factor to obtain an improved Harris Hawk algorithm.

It should be explained that step S02 includes:

Step S021: Introducing Chebyshev chaotic mapping into the global exploration phase of the Harris Hawk algorithm to obtain an improved global exploration phase.

It should be noted that the Chebyshev chaotic map is:

CM＝cos(tcos ^-1 (γ)) (1)

Where t is the current iteration number and γ is the initial value of chaos.

The improved model of the global exploration stage is:

Where X(t+1) and X(t) are the next iteration position and current position of Harris's hawk, respectively; X _rabbit (t) is the location of the prey; CM ₁ , CM ₂ , CM ₃ , CM ₄ , and q are all random numbers between (0,1); UB and LB are the upper and lower limits of the individual position search, respectively; X _m (t) is the average position of the population;

Step S022: using a nonlinear decreasing disturbance factor to improve the prey escape energy in the Harris Hawk algorithm to obtain an improved escape energy formula.

It should be explained that the improved escape energy formula is:

E＝2E ₀ (2r(0.5+cos((π(t/T)) ^1/2 ))) (3)

Where E _{= 0} , the initial escape energy, T = the maximum number of iterations, t = the current number of iterations, and r = a random number between (0, 1).

Step S023: weighting the optimal individual in the population in the Harris Hawk algorithm using an adaptive weight factor to obtain an optimal individual update formula.

It should be noted that the adaptive weight factor is:

The optimal individual update formula:

X′ _rabbit = w(t) × X _rabbit (5)

Step S024: obtaining an improved Harris Hawk algorithm based on the improved global exploration phase, the improved escape energy formula and the optimal individual update formula.

It needs to be explained that the improved Harris Hawk algorithm is consistent with the traditional Harris Hawk algorithm in the local development stage.

In the local development stage: Harris hawk populations can use four different strategies to update the population position according to the escape behavior of prey, and choose the siege strategy according to the escape energy E and the hunting random selection factor r∈(0,1).

1) When r is greater than or equal to 0.5 and |E| is greater than or equal to 0.5, the Harris Hawk implements a soft siege strategy, and the model is as follows:

X(t+1)＝ΔX(t)-E|JX _rabbit (t)-X(t)| (6)

ΔX(t)＝ _Xrabbit (t)-X(t) (7)

J＝2(1-r ₅ ) (8)

Among them, J is the random jumping intensity of the prey during the entire escape process, and X _rabbit (t) is the optimal solution in the current population.

2) When r is greater than or equal to 0.5 and |E| is less than 0.5, Harris's hawk uses a hard siege strategy to capture the prey directly. The model is as follows:

X(t+1)＝ _Xrabbit (t)-E|ΔX(t)| (9)

3) When r is less than 0.5 and |E| is greater than or equal to 0.5, the Harris Hawk implements a gradual rapid dive soft siege strategy, and the model is as follows:

Where D is the dimension of the problem to be optimized, S is a D-dimensional random vector, and LF is the Levy flight operator.

4) When r is less than 0.5 and |E| is less than 0.5, the Harris Hawk implements a progressive fast dive hard siege strategy, and the model is as follows:

For example, referring to FIG4, FIG4 is a schematic diagram of the improved Harris Hawk algorithm flow in the second-hand car price prediction method of the present invention. First, the HHO parameters are initialized, the population is randomly initialized, and the individual fitness is calculated; the escape energy E is updated according to formula (3); then it is judged that E>1?; if E>1, the position is updated according to formula (1) and formula (2), and then t>T? is judged. If t>T, the fitness and the optimal individual are output. If t≤T, E>1? is judged again. If E≤1, when r is greater than or equal to 0.5 and |E| is greater than or equal to 0.5, the position is updated according to formula (6); when r is greater than or equal to 0.5 and |E| is less than 0.5, the position is updated according to formula (9); when r is less than 0.5 and |E| is greater than or equal to 0.5, the position is updated according to formula (4), (5) and (10); when r is less than 0.5 and |E| is less than 0.5, the position is updated according to formula (4), (5) and (11), and finally t>T? is judged. , if t>T, then output the fitness and the optimal individual.

Step S03: Use the improved Harris Eagle algorithm to optimize the hyperparameters of the LightGBM model to obtain the target parameter combination.

It should be explained that the hyperparameters of the LightGBM model in this embodiment and the following embodiments may be the number of leaf nodes of each tree in the LightGBM model, the maximum depth of each tree, the model learning rate, and the minimum number of samples required for each leaf node.

In a specific implementation, the population size, number of iterations and problem dimension of the improved Harris Hawk algorithm can be set to obtain the target Harris Hawk algorithm; the number of leaf nodes of each tree, the maximum depth of each tree, the model learning rate and the minimum number of samples required for each leaf node in the LightGBM model are used as hyperparameters of the LightGBM model; the training data set is input into the LightGBM model, and the individual fitness value is determined according to a preset fitness function; based on the individual fitness value, the population individuals are iteratively updated according to the target Harris Hawk algorithm until the number of iterations is reached to obtain the target parameter combination.

It should be noted that the above preset fitness function is:

Where _ypred is the predicted value of the used car price corresponding to sample i, _yi is the actual value of the used car price of sample i, n is the number of samples, and parameter is the hyperparameter of the optimized LightGBM model.

Step S04: After establishing an initial LightGBM model based on the target parameter combination, the initial LightGBM model is trained using the training data set to obtain a target used car price prediction model.

In a specific implementation, an initial LightGBM model can be established based on the LightGBM model and the target parameter combination; the initial LightGBM model is trained using the training data set to obtain a training result; the initial LightGBM model is optimized according to the training result to obtain a target used car price prediction model.

This embodiment obtains a used car sales data set, and pre-processes the used car sales data set to obtain a training data set; improves the Harris Hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to obtain an improved Harris Hawk algorithm; uses the improved Harris Hawk algorithm to optimize the hyperparameters of the LightGBM model to obtain a target parameter combination; after establishing an initial LightGBM model based on the target parameter combination, the initial LightGBM model is trained using the training data set to obtain a target used car price prediction model. Compared with the prior art, this embodiment improves the Harris Hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to obtain an improved Harris Hawk algorithm, and then uses the improved Harris Hawk algorithm to optimize the hyperparameters of the LightGBM model, thereby solving the problem that the subjective parameter setting of the LightGBM model is prone to fall into the local optimum, and improving the learning ability of the LightGBM model.

Refer to FIG5 , which is a flow chart of a third embodiment of a method for predicting used car prices according to the present invention.

Based on the above embodiments, in this embodiment, step S01 includes:

Step S011: obtaining a used car sales dataset, and preprocessing the used car sales dataset to obtain a preprocessed used car sales dataset.

Step S012: split the preprocessed used car sales data set according to a preset ratio to obtain a training data set and a test data set.

It is understandable that the above preset ratio may be a user-defined setting or a specific setting according to the used car sales data set, for example, 6:4, 7:3 or 8:2, etc., which is not limited in this embodiment.

In a specific implementation of this embodiment, the preprocessed used car sales data set may be split in a ratio of 8:2 to obtain a training data set and a test data set, which are used for model training and testing, respectively.

Accordingly, after the step of training the initial LightGBM model using the training data set to obtain the target used car price prediction model, the method further includes:

Step S05: using the target used car price prediction model to predict the test data set to obtain a used car price prediction value.

Step S06: selecting evaluation indicators, and evaluating the target used car price prediction model based on the used car price prediction value, wherein the evaluation indicators include determination coefficient, mean square error and mean absolute error.

For example, referring to FIG. 6 , FIG. 6 is a schematic diagram of the overall design of the target used car price prediction model in the used car price prediction method of the present invention.

Step 1: Preprocess the data set and split the data set into a training set and a test set (i.e., split the preprocessed used car sales data set according to a preset ratio to obtain a training data set and a test data set);

Step 2: Set the population size N, number of iterations T and problem dimension D of the iHHO algorithm (i.e. the improved Harris Hawk algorithm);

Step 3: Set the search range of four LightGBM model hyperparameters, including the number of leaf nodes per tree num_leaves, the maximum depth of each tree max_depth, the model learning rate learning_rate, and the minimum number of samples required for each leaf node min_child_samples;

Step 4: Input the training set (i.e., training data set) into the LightGBM model and calculate the individual fitness value according to formula (12);

Step 5: Iteratively update the population individuals according to the iHHO algorithm;

Step 6: If the iteration termination condition is reached, go to Step 7; otherwise, return to Step 4;

Step 7: Output the optimal parameter combination (i.e., target parameter combination);

Step 8: Establish the LightGBM model based on the optimal parameter combination, and train the model using the training set (training data set);

Step 9: Use the trained model to predict the test set;

Step 10: Output the predicted value of used car prices and evaluate the prediction model.

It should be noted that the above determination coefficient R ² , mean square error RMSE and mean absolute error MAE are defined as follows:

Where n is the number of samples, y _pred is the predicted value of the used car price, y _mean is the mean, and y is the true value of the used car price. MAE measures the overall prediction accuracy, and the smaller the value, the better the data fit. RMSE quantifies the size of the prediction error, and a smaller value indicates a higher prediction accuracy. R ² measures the degree to which the model fits the variability of the target variable, and a value closer to 1 indicates a better fit.

In the specific implementation, the open used car dataset VW can also be introduced to verify the target used car price prediction model, and the Pearson correlation coefficient can be used to perform a correlation analysis on the characteristics that affect the used car price, proving that the target used car price prediction model is superior to the comparison model in terms of prediction accuracy and data fitting.

This embodiment describes the step of using the training data set to train the initial LightGBM model to obtain the target used car price prediction model, and then uses the target used car price prediction model to predict the test data set to obtain the used car price prediction value; selects evaluation indicators, and evaluates the target used car price prediction model based on the used car price prediction value, and the evaluation indicators include determination coefficient, mean square error and mean absolute error. Compared with the prior art, since this embodiment selects determination coefficient, mean square error and mean absolute error to evaluate the target used car price prediction model, the evaluation of the target used car price prediction model is achieved, proving that the target used car price prediction model is superior to the comparison model in terms of prediction accuracy and data fitting.

In addition, an embodiment of the present invention further proposes a storage medium, on which a used car price prediction program is stored. When the used car price prediction program is executed by a processor, the steps of the used car price prediction method described above are implemented.

Refer to FIG. 7 , which is a structural block diagram of a first embodiment of a second-hand car price prediction device according to the present invention.

As shown in FIG. 7 , the second-hand car price prediction device provided in the embodiment of the present invention includes: a data processing module 701 and a price output module 702 .

The data processing module 701 is used to obtain vehicle attribute information of a used car, and pre-process the vehicle attribute information to obtain target used car data.

The price output module 702 is used to input the target used car data into a target used car price prediction model to obtain a used car price prediction value. The target used car price prediction model is a model obtained by improving the Harris Eagle algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor, and optimizing the hyperparameters of the LightGBM model based on the improved Harris Eagle algorithm.

This embodiment obtains target second-hand car data by acquiring vehicle attribute information of used cars and preprocessing the vehicle attribute information; the target second-hand car data is input into a target second-hand car price prediction model to obtain a second-hand car price prediction value, wherein the target second-hand car price prediction model is a model obtained by optimizing the hyperparameters of the LightGBM model based on the improved Harris Hawk algorithm after introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to improve the Harris Hawk algorithm's global exploration and local development capabilities, thereby improving the algorithm's global search accuracy, and then optimizing the hyperparameters of LightGBM based on the improved Harris Hawk algorithm, thereby solving the hyperparameter dependency problem in model prediction, thereby effectively improving the accuracy and stability of second-hand car price prediction.

Based on the first embodiment of the second-hand car price prediction device of the present invention, a second embodiment of the second-hand car price prediction device of the present invention is proposed.

In this embodiment, the used car price prediction module 701 is also used to obtain a used car sales data set, and preprocess the used car sales data set to obtain a training data set; improve the Harris Hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to obtain an improved Harris Hawk algorithm; use the improved Harris Hawk algorithm to optimize the hyperparameters of the LightGBM model to obtain a target parameter combination; after establishing an initial LightGBM model based on the target parameter combination, use the training data set to train the initial LightGBM model to obtain a target used car price prediction model.

The used car price prediction module 701 is also used to introduce Chebyshev chaotic mapping in the global exploration stage of the Harris Hawk algorithm to obtain an improved global exploration stage; use a nonlinear decreasing disturbance factor to improve the prey escape energy in the Harris Hawk algorithm to obtain an improved escape energy formula; use an adaptive weight factor to weight the optimal individual in the population in the Harris Hawk algorithm to obtain an optimal individual update formula; and obtain an improved Harris Hawk algorithm based on the improved global exploration stage, the improved escape energy formula and the optimal individual update formula.

The used car price prediction module 701 is also used to set the population size, number of iterations and problem dimension of the improved Harris Hawk algorithm to obtain the target Harris Hawk algorithm; the number of leaf nodes of each tree, the maximum depth of each tree, the model learning rate and the minimum number of samples required for each leaf node in the LightGBM model are used as hyperparameters of the LightGBM model; the training data set is input into the LightGBM model, and the individual fitness value is determined according to a preset fitness function; based on the individual fitness value, the population individuals are iteratively updated according to the target Harris Hawk algorithm until the number of iterations is reached to obtain the target parameter combination.

The used car price prediction module 701 is also used to establish an initial LightGBM model based on the LightGBM model and the target parameter combination; train the initial LightGBM model using the training data set to obtain a training result; optimize the initial LightGBM model according to the training result to obtain a target used car price prediction model.

This embodiment obtains a used car sales data set, and pre-processes the used car sales data set to obtain a training data set; improves the Harris Hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to obtain an improved Harris Hawk algorithm; uses the improved Harris Hawk algorithm to optimize the hyperparameters of the LightGBM model to obtain a target parameter combination; after establishing an initial LightGBM model based on the target parameter combination, the initial LightGBM model is trained using the training data set to obtain a target used car price prediction model. Compared with the prior art, this embodiment improves the Harris Hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factor and adaptive weight factor to obtain an improved Harris Hawk algorithm, and then uses the improved Harris Hawk algorithm to optimize the hyperparameters of the LightGBM model, thereby solving the problem that the subjective parameter setting of the LightGBM model is prone to fall into the local optimum, and improving the learning ability of the LightGBM model.

Other embodiments or specific implementations of the second-hand car price prediction device of the present invention can refer to the above-mentioned method embodiments and will not be described in detail here.

It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory/random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims (10)

1. The second-hand vehicle price prediction method is characterized by comprising the following steps of:

Acquiring second-hand vehicle attribute information, and preprocessing the vehicle attribute information to acquire target second-hand vehicle data;

And inputting the target second-hand vehicle data into a target second-hand vehicle price prediction model to obtain a second-hand vehicle price prediction value, wherein the target second-hand vehicle price prediction model is obtained by introducing Chebyshev chaotic mapping, nonlinear decreasing disturbance factors and self-adaptive weight factors to improve a Harris eagle algorithm and optimizing super-parameters of a LightGBM model based on the improved Harris eagle algorithm.

2. The second-hand-car price prediction method according to claim 1, wherein the step of acquiring second-hand-car vehicle attribute information and preprocessing the vehicle attribute information to obtain target second-hand-car data further comprises, before:

Acquiring a second-hand vehicle sales data set, and preprocessing the second-hand vehicle sales data set to acquire a training data set;

improving the Harris hawk algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing disturbance factors and self-adaptive weight factors, so as to obtain an improved Harris hawk algorithm;

Optimizing the super parameters of the LightGBM model by utilizing the improved Harris eagle algorithm to obtain a target parameter combination;

and after an initial LightGBM model is established based on the target parameter combination, training the initial LightGBM model by utilizing the training data set to obtain a target secondary car price prediction model.

3. The second-hand car price prediction method of claim 2, wherein the step of improving the harris eagle algorithm by introducing Chebyshev chaotic mapping, nonlinear decreasing perturbation factors and adaptive weighting factors, and obtaining the improved harris eagle algorithm comprises the following steps:

Introducing Chebyshev chaotic mapping into the global exploration phase of the Harris hawk algorithm to obtain an improved global exploration phase;

Improving the escaping energy of the prey in the Harris eagle algorithm by adopting a nonlinear decreasing disturbance factor to obtain an improved escaping energy formula;

Weighting the optimal individuals in the population in the Harris eagle algorithm by using the self-adaptive weight factors to obtain an optimal individual updating formula;

and obtaining an improved Harris hawk algorithm based on the improved global exploration phase, the improved escape energy formula and the optimal individual update formula.

4. The second-hand-car price prediction method of claim 3, wherein the Chebyshev chaotic map is:

CM＝cos(tcos^-1(γ))

Wherein t is the current iteration number, and gamma is the chaos initial value;

the improved global exploration phase model is as follows:

wherein X (t+1) and X (t) are respectively the position and the current position of the next iteration of the Harris eagle, X _rabbit (t) is the position of the prey, CM ₁、CM₂、CM₃、CM₄ and q are random numbers between (0 and 1), UB and LB are respectively the upper limit and the lower limit of individual position search, X _m (t) is the average position of the population,

Accordingly: the improved escape energy formula is as follows:

E＝2E₀(2r(0.5+cos((π(t/T))^1/2)))；

Wherein E ₀ is initial escape energy, T is maximum iteration times, T is current iteration times, and r is a random number between (0, 1);

The adaptive weight factor is:

The optimal individual update formula:

X′_rabbit＝w(t)×X_rabbit。

5. The second-hand-car price prediction method of claim 2, wherein the step of acquiring a second-hand-car sales data set and preprocessing the second-hand-car sales data set to obtain a training data set comprises:

acquiring a second-hand vehicle sales data set, preprocessing the second-hand vehicle sales data set, and acquiring a preprocessed second-hand vehicle sales data set;

splitting the preprocessed second-hand vehicle sales data set according to a preset proportion to obtain a training data set and a testing data set;

Correspondingly, after the step of training the initial LightGBM model by using the training dataset to obtain the target secondary handcart price prediction model, the method further includes:

predicting the test data set by using the target secondary vehicle price prediction model to obtain a secondary vehicle price prediction value;

And selecting an evaluation index, and evaluating the target secondary car price prediction model based on the secondary car price prediction value, wherein the evaluation index comprises a decision coefficient, a mean square error and an average absolute error.

6. The second-hand-car price prediction method of claim 2, wherein the step of optimizing the super-parameters of the LightGBM model by using the modified harris eagle algorithm to obtain the target parameter combination comprises the following steps:

setting the population scale, iteration times and problem dimension of the improved Harris eagle algorithm to obtain a target Harris eagle algorithm;

Taking the number of leaf nodes in the LightGBM model, the maximum depth of each tree, the model learning rate and the minimum number of samples required by each leaf node as super parameters of the LightGBM model;

inputting the training data set into the LightGBM model, and determining an individual fitness value according to a preset fitness function;

And based on the individual fitness value, carrying out iterative updating on population individuals according to the target Harriset algorithm until the iterative times are reached, and obtaining a target parameter combination.

7. The method of claim 6, wherein the step of training the initial LightGBM model with the training dataset after establishing the initial LightGBM model based on the target parameter combination to obtain a target secondary price prediction model comprises:

Establishing an initial LightGBM model based on the LightGBM model and the target parameter combination;

Training the initial LightGBM model by using the training data set to obtain a training result;

and optimizing the initial LightGBM model according to the training result to obtain a target secondary handcart price prediction model.

8. A secondary car price prediction apparatus, the apparatus comprising:

the data processing module is used for acquiring the attribute information of the second-hand vehicle, preprocessing the attribute information of the second-hand vehicle and acquiring target second-hand vehicle data;

the price output module is used for inputting the target second-hand vehicle data into a target second-hand vehicle price prediction model to obtain a second-hand vehicle price prediction value, and the target second-hand vehicle price prediction model is obtained by introducing Chebyshev chaotic mapping, nonlinear decreasing disturbance factors and self-adaptive weight factors to improve a Harris eagle algorithm and optimizing super-parameters of a LightGBM model based on the improved Harris eagle algorithm.

9. A secondary car price prediction apparatus, the apparatus comprising: a memory, a processor and a second-hand-vehicle price prediction program stored on the memory and executable on the processor, the second-hand-vehicle price prediction program being configured to implement the steps of the second-hand-vehicle price prediction method of any one of claims 1 to 7.

10. A storage medium having stored thereon a second-hand-vehicle price prediction program which, when executed by a processor, implements the steps of the second-hand-vehicle price prediction method according to any one of claims 1 to 7.