Second-hand car price prediction method, system and equipment based on multi-model fusion
Through a multi-model fusion method, combined with the weighted fusion and parameter optimization of historical used car transaction data and multiple prediction models, the problems of insufficient accuracy, interpretability and robustness of the existing used car price prediction models are solved, and more efficient used car price prediction is achieved.
Patent Information
- Application Number
- CN202411983618.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing used car price prediction models generally rely only on the vehicle's characteristics and historical transaction data for prediction, with low accuracy, low interpretability of the prediction model and poor robustness.
A multi-model fusion method is adopted to obtain historical used car transaction data for preprocessing, and a variety of prediction models (such as MLP model and index system method model) are established, and weighted fusion and parameter optimization are carried out to obtain a more accurate and interpretable used car price prediction model.
It improves the accuracy, interpretability and robustness of used car price predictions, providing a more accurate and trustworthy prediction method.
Smart Images
Figure CN119904283A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of financial management technology, and in particular to a method, system and device for predicting used car prices based on multi-model fusion. Background Art
[0002] With the rapid development of the automobile industry, the used car market has gradually matured worldwide and has become an important part of the automobile industry chain. The active used car trading market provides consumers with a variety of choices and has become the main way for many families to buy cars. However, the price evaluation of used cars is a complex process involving multiple factors, such as vehicle brand, model, age, mileage, vehicle status, market demand, etc. Therefore, how to accurately evaluate the value of used cars has become an important topic in used car transactions.
[0003] Traditionally, the price evaluation of used cars mainly relies on manual evaluation and experience judgment. The evaluator gives a valuation by checking the appearance, performance, maintenance history and other aspects of the vehicle, combined with market conditions and transaction data. This manual evaluation method has certain limitations, which are mainly reflected in the following aspects: first, the evaluation results are greatly affected by the experience and subjective factors of the evaluator, which may lead to valuation errors; second, due to the frequent changes in market conditions and vehicle conditions, traditional manual evaluation often lacks real-time and accuracy; finally, the manual evaluation process is cumbersome and time-consuming, which affects the efficiency of used car transactions. With the development of information technology, used car valuation technologies based on big data and artificial intelligence have gradually emerged. These technologies can realize the automated and intelligent evaluation of used cars by analyzing a large amount of historical transaction data, vehicle information and market dynamics. For example, through machine learning algorithms, various attributes of the vehicle can be modeled, and a valuation model can be trained based on historical transaction data, thereby realizing the automatic valuation of used cars. Compared with traditional manual evaluation methods, this method can provide more accurate and fair price evaluation in a shorter time.
[0004] However, existing used car price prediction models generally only rely on vehicle characteristics and historical transaction data for prediction, which has low accuracy, low interpretability and poor robustness. Summary of the invention
[0005] The present invention provides a used car price prediction method, system and device based on multi-model fusion, which at least solves the problem that the existing used car price prediction model generally only relies on the characteristics of the vehicle and historical transaction data for prediction, has low accuracy, low interpretability of the prediction model, and poor robustness, and provides a used car price prediction method with high accuracy, high interpretability and strong robustness.
[0006] A used car price prediction method based on multi-model fusion, including:
[0007] Acquire historical second-hand car transaction data, and pre-process the historical second-hand car transaction data;
[0008] According to the pre-processed historical used car transaction data, an MLP model is established and trained to obtain the first value prediction model;
[0009] According to the pre-processed historical second-hand car transaction data, the second value prediction model is obtained based on the indicator system method;
[0010] Performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model;
[0011] According to the preprocessed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain a second-hand car value prediction model;
[0012] The information of the vehicle to be evaluated is input into the second-hand car value prediction model to obtain the second-hand car price of the vehicle to be evaluated.
[0013] Optionally, establishing an MLP model and training the MLP model according to the preprocessed historical used car transaction data to obtain a first value prediction model includes:
[0014] Establishing an MLP model based on the preprocessed historical used car transaction data, wherein the MLP model is configured to include an input layer, a hidden layer, and an output layer;
[0015] According to the preprocessed historical used car transaction data, a training set and a test set are obtained;
[0016] The MLP model is trained according to the training set and the test set to obtain a first value prediction model.
[0017] Optionally, obtaining the second value prediction model based on the preprocessed historical used car transaction data and the indicator system method includes:
[0018] The method of obtaining key indicators in the historical second-hand car transaction data based on the pre-processed historical second-hand car transaction data;
[0019] According to the key indicators in the historical used car transaction data, based on the indicator system method, a used car value indicator system is obtained;
[0020] According to the second-hand car value index system, a second-hand car value prediction rule is obtained;
[0021] According to the second-hand car value prediction rule, a second value prediction model is obtained.
[0022] Optionally, performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model includes:
[0023] According to the first value prediction model, obtaining a first prediction result of the first value prediction model;
[0024] According to the second value prediction model, obtaining a second prediction result of the second value prediction model;
[0025] A third value prediction model is obtained by weighted fusion of the first prediction result and the second prediction result. The third value prediction model is configured as follows:
[0026] The third prediction result = a*first prediction result + b*second prediction result;
[0027] Among them, the input of the third value prediction model is the first prediction result and the second prediction result, the output of the third value prediction model is the third prediction result, a is the weight coefficient of the first value prediction model, and b is the weight coefficient of the second value prediction model.
[0028] Optionally, the step of optimizing weighted parameters of the third value prediction model based on the preprocessed historical used car transaction data to obtain a used car value prediction model includes:
[0029] According to the pre-processed historical used car transaction data and the third value prediction model, a first set of weight coefficients is obtained based on the grid search method;
[0030] According to the pre-processed historical used car transaction data and the third value prediction model, a second set of weight coefficients is obtained based on the gradient descent method;
[0031] Based on the first set of weight coefficients and / or the second set of weight coefficients, weighted parameters of the third value prediction model are optimized to obtain a used car value prediction model.
[0032] Optionally, also include:
[0033] Establishing a value correction model, wherein the value correction model is configured to take a vehicle model and an event associated with the vehicle model as input and take a price change as output;
[0034] The method of optimizing the weighted parameters of the third value prediction model based on the preprocessed historical second-hand car transaction data to obtain the second-hand car value prediction model includes:
[0035] The weighted parameter optimization of the third value prediction model is performed according to the preprocessed historical second-hand car transaction data to obtain an optimized third value prediction model;
[0036] A used car value prediction model is obtained based on the optimized third value prediction model and the value correction model.
[0037] Optionally, the establishing of the value correction model includes:
[0038] Establishing an initial value correction model, wherein the initial value correction model is configured as a regression analysis model, with the vehicle model and the associated events of the vehicle model as inputs and the price change as output;
[0039] Based on historical second-hand car transaction data, obtain the transaction price distribution of the same second-hand cars with similar vehicle characteristics;
[0040] According to the price distribution, obtaining at least one price revision data;
[0041] According to the price correction data, matching the associated events corresponding to the price correction data;
[0042] The regression analysis model is trained according to the price correction data, the associated events corresponding to the price correction data, and the vehicle model corresponding to the price correction data to obtain a value correction model.
[0043] Optionally, obtaining the transaction price distribution of the same used cars with similar vehicle characteristics based on historical used car transaction data includes:
[0044] According to the historical second-hand car transaction data, the data is classified according to the car model, and the historical second-hand car transaction data of at least one car model is obtained;
[0045] Clustering the historical second-hand car transaction data of each model according to vehicle characteristics to obtain the first clustering result;
[0046] According to the first clustering result, the transaction price and transaction time of each transaction in each cluster are obtained;
[0047] Clustering the data of each cluster in the first clustering result again according to the transaction price and transaction time to obtain a second clustering result;
[0048] According to the second clustering result, the transaction price distribution of the same used cars with similar characteristics is obtained.
[0049] In another aspect, a used car price prediction system based on multi-model fusion includes a transaction data analysis platform and a transaction data storage platform;
[0050] The transaction data analysis platform is configured to:
[0051] Acquire historical second-hand car transaction data, and pre-process the historical second-hand car transaction data;
[0052] According to the pre-processed historical used car transaction data, an MLP model is established and trained to obtain the first value prediction model;
[0053] According to the pre-processed historical second-hand car transaction data, the second value prediction model is obtained based on the indicator system method;
[0054] Performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model;
[0055] According to the preprocessed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain a second-hand car value prediction model;
[0056] Inputting the information of the vehicle to be evaluated into the second-hand car value prediction model to obtain the second-hand car price of the vehicle to be evaluated;
[0057] The transaction data storage platform is configured as follows:
[0058] The historical second-hand car transaction data is stored, and in response to a request from the transaction data analysis platform, the historical second-hand car transaction data is sent to the transaction data analysis platform.
[0059] On the other hand, a computer device is provided. The computer device includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement the above method.
[0060] On the other hand, a computer storage medium stores a computer program, and a processor executes the computer program to implement the above method.
[0061] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0062] The present invention provides a second-hand car price prediction method, system and device based on multi-model fusion, including: obtaining historical second-hand car transaction data, preprocessing the historical second-hand car transaction data; establishing an MLP model and training the MLP model according to the preprocessed historical second-hand car transaction data to obtain a first value prediction model; obtaining a second value prediction model based on the index system method according to the preprocessed historical second-hand car transaction data; performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model; optimizing the weighted parameters of the third value prediction model according to the preprocessed historical second-hand car transaction data to obtain a second-hand car value prediction model; inputting the information of the vehicle to be evaluated into the second-hand car value prediction model to obtain the second-hand car price of the vehicle to be evaluated. At least the problem that the existing second-hand car price prediction model generally only relies on the characteristics of the vehicle and the historical transaction data for prediction, has low accuracy, low interpretability of the prediction model, and poor robustness is solved, and a second-hand car price prediction method with high accuracy, high interpretability and strong robustness is provided. BRIEF DESCRIPTION OF THE DRAWINGS
[0063] In order to more clearly illustrate the specific implementation of the present application or the technical solution in the prior art, the following is a brief introduction to the drawings required for the specific implementation or the prior art description. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.
[0064] Figure 1 A flowchart of a method for predicting second-hand car prices based on multi-model fusion in this application;
[0065] Figure 2 This is a schematic diagram of the structure of a computer device in this application.
[0066] Markings in the figure: 101 - processor, 102 - communication bus, 103 - network interface, 104 - user interface, 105 - memory.
[0067] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0068] In order to enable those skilled in the art to better understand the scheme of the present disclosure, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present disclosure.
[0069] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. A first user refers to one or more users that meet the first user characteristics, and a second user refers to one or more users that meet the second user characteristics.
[0070] Example 1
[0071] A used car price prediction method based on multi-model fusion, including:
[0072] S1. Obtain historical second-hand car transaction data and pre-process the historical second-hand car transaction data.
[0073] Optionally, each piece of used car transaction data in the historical used car transaction data includes one or more of the transaction number, transaction date, vehicle model, vehicle characteristics and transaction price, wherein the vehicle characteristics may include vehicle condition data such as vehicle age, vehicle maintenance condition, and mileage.
[0074] Optionally, a preprocessing method for preprocessing historical used car transaction data may include: data cleaning processing and feature engineering processing.
[0075] S2. Based on the preprocessed historical used car transaction data, an MLP model is established and trained to obtain a first value prediction model.
[0076] Optionally, the MLP model is a feedforward neural network model, which consists of an input layer, a hidden layer, and an output layer. The first value prediction model can be obtained by training the MLP model using historical used car transaction data.
[0077] Optionally, the architecture design of the MLP model includes an input layer, a hidden layer, and an output layer, where:
[0078] The input layer is configured to input features in the used car transaction data, such as vehicle model, vehicle features, etc.
[0079] The hidden layer is configured as a multi-layer, using the ReLU activation function;
[0080] The output layer is configured as a single regression output node that outputs the predicted price A.
[0081] Optionally, the process of training the MLP model includes:
[0082] Use historical used car transaction data to train the model;
[0083] Use a suitable loss function, such as MSE, to measure the difference between the predicted result and the actual price;
[0084] Choose an optimization algorithm, such as the Adam optimizer, to update the model weights.
[0085] S3. According to the pre-processed historical second-hand car transaction data, based on the indicator system method, a second value prediction model is obtained.
[0086] Optionally, a set of value assessment rules can be obtained through the indicator system method, and the price B can be estimated through the value assessment rules. For example, weights can be set based on factors such as vehicle age, mileage, and vehicle condition, and the price can be calculated according to the rules; for example, an example rule:
[0087] Depreciation is 5% per year based on the age of the vehicle;
[0088] Based on mileage, a 2% discount will be given for every additional 10,000 kilometers;
[0089] Discounts or price increases will be given based on the condition of the vehicle, such as accident vehicle or non-accident vehicle.
[0090] S4. Perform weighted fusion on the first value prediction model and the second value prediction model to obtain a third value prediction model.
[0091] Optionally, the price A predicted by the MLP model and the price B predicted by the rule are weighted and fused, and the final price prediction C is obtained by a weighted method. Specifically, C=aA+bB, where a and b are weight coefficients. The goal of the third value prediction model is to determine these two weight coefficients so that the fused result C is closest to the actual transaction price.
[0092] S5. Optimize the weighted parameters of the third value prediction model based on the preprocessed historical second-hand car transaction data to obtain a second-hand car value prediction model.
[0093] Based on historical data, the weight coefficient can be calculated by linear regression or other regression methods, where the input is configured as the actual price and price A and price B in the historical used car transaction data; the output is configured with weight coefficients a and b. In order to optimize the final price prediction C, it is necessary to select the optimal weight coefficients a and b, which can be optimized by grid search method or gradient descent method.
[0094] S6. Input the information of the vehicle to be evaluated into a used car value prediction model to obtain the used car price of the vehicle to be evaluated.
[0095] Optionally, the information of the vehicle to be evaluated is input into a used car value prediction model to obtain the used car value of the vehicle to be evaluated, and points are added or evaluated based on the used car value to obtain the used car price of the vehicle to be evaluated.
[0096] This solution introduces an indicator system generated by the experience of traditional second-hand car traders, integrates the traditional second-hand car value analysis method with the AI second-hand car value analysis method, and improves the prediction accuracy, interpretability and robustness of the entire prediction method. At least it solves the problem that the existing second-hand car price prediction model generally only relies on the characteristics of the vehicle and historical transaction data for prediction, with low accuracy, low interpretability of the prediction model, and poor robustness, and provides a second-hand car price prediction method with high accuracy, high interpretability and strong robustness.
[0097] Example 2
[0098] This embodiment is based on the first embodiment, and is a method for predicting second-hand car prices based on multi-model fusion, including:
[0099] S1. Obtain historical second-hand car transaction data and pre-process the historical second-hand car transaction data.
[0100] Optionally, each piece of used car transaction data in the historical used car transaction data includes one or more of the transaction number, transaction date, vehicle model, vehicle characteristics and transaction price, wherein the vehicle characteristics may include vehicle condition data such as vehicle age, vehicle maintenance condition, and mileage.
[0101] Optionally, a preprocessing method for preprocessing historical used car transaction data may include: data cleaning processing and feature engineering processing.
[0102] S2. Based on the preprocessed historical used car transaction data, an MLP model is established and trained to obtain a first value prediction model.
[0103] Optionally, the MLP model is a feedforward neural network model, which consists of an input layer, a hidden layer, and an output layer. The first value prediction model can be obtained by training the MLP model using historical used car transaction data.
[0104] Optionally, the architecture design of the MLP model includes an input layer, a hidden layer, and an output layer, where:
[0105] The input layer is configured to input features in the used car transaction data, such as vehicle model, vehicle features, etc.
[0106] The hidden layer is configured as a multi-layer, using the ReLU activation function;
[0107] The output layer is configured as a single regression output node that outputs the predicted price A.
[0108] Optionally, the process of training the MLP model includes:
[0109] Use historical used car transaction data to train the model;
[0110] Use a suitable loss function, such as MSE, to measure the difference between the predicted result and the actual price;
[0111] Choose an optimization algorithm, such as the Adam optimizer, to update the model weights.
[0112] Optionally, based on the preprocessed historical used car transaction data, an MLP model is established and trained to obtain a first value prediction model, including:
[0113] According to the preprocessed historical used car transaction data, an MLP model is established, and the MLP model is configured to include an input layer, a hidden layer, and an output layer;
[0114] According to the preprocessed historical used car transaction data, a training set and a test set are obtained;
[0115] The MLP model is trained according to the training set and the test set to obtain the first value prediction model.
[0116] Specifically, a Python code example that implements the above steps:
[0117] from sklearn.neural_network import MLPRegressor
[0118] from sklearn.pipeline import Pipeline
[0119] #Create MLP model
[0120] mlp_model=Pipeline(steps=[
[0121] ('preprocessor',preprocessor),#data preprocessing
[0122] ('mlp',MLPRegressor(hidden_layer_sizes=(100,),max_iter=1000,random_state=42))#MLP regression model ])
[0124] #Train model
[0125] mlp_model.fit(X_train,y_train)
[0126] #Predicted price
[0127] y_pred=mlp_model.predict(X_test)
[0128] #Print prediction results
[0129] print("Predicted price:",y_pred)
[0130] print("Real price:",y_test.values).
[0131] Among them, hidden_layer_sizes = (100,) defines a hidden layer with 100 neurons in the layer. This parameter can be adjusted to change the network structure;
[0132] max_iter=1000 means the maximum number of iterations is 1000;
[0133] random_state=42: Set the random seed to ensure repeatable results.
[0134] S3. According to the pre-processed historical second-hand car transaction data, based on the indicator system method, a second value prediction model is obtained.
[0135] Optionally, a set of value assessment rules can be obtained through the indicator system method, and the price B can be estimated through the value assessment rules. For example, weights can be set based on factors such as vehicle age, mileage, and vehicle condition, and the price can be calculated according to the rules; for example, an example rule:
[0136] Depreciation is 5% per year based on the age of the vehicle;
[0137] Based on mileage, a 2% discount will be given for every additional 10,000 kilometers;
[0138] Discounts or price increases will be given based on the condition of the vehicle, such as accident vehicle or non-accident vehicle.
[0139] Optionally, according to the pre-processed historical used car transaction data, based on the indicator system method, a second value prediction model is obtained, including:
[0140] According to the pre-processed historical second-hand car transaction data, key indicators in the historical second-hand car transaction data are obtained;
[0141] According to the key indicators in the historical second-hand car transaction data, based on the indicator system method, the second-hand car value indicator system is obtained;
[0142] According to the second-hand car value index system, the second-hand car value prediction rules are obtained;
[0143] According to the used car value prediction rules, a second value prediction model is obtained.
[0144] Optionally, according to key indicators in historical used car transaction data, based on the indicator system method, methods for obtaining a used car value indicator system include expert method, regression analysis method, hierarchical analysis method or hybrid method.
[0145] Optionally, according to the used car value index system, a method for obtaining used car value prediction rules includes extracting rules based on expert experience, extracting rules based on data mining, and generating rules based on fuzzy logic.
[0146] S4. Perform weighted fusion on the first value prediction model and the second value prediction model to obtain a third value prediction model.
[0147] Optionally, the price A predicted by the MLP model and the price B predicted by the rule are weighted and fused, and the final price prediction C is obtained by a weighted method. Specifically, C=aA+bB, where a and b are weight coefficients. The goal of the third value prediction model is to determine these two weight coefficients so that the fused result C is closest to the actual transaction price.
[0148] S5. Optimize the weighted parameters of the third value prediction model based on the preprocessed historical second-hand car transaction data to obtain a second-hand car value prediction model.
[0149] Based on historical data, the weight coefficient can be calculated by linear regression or other regression methods, where the input is configured as the actual price and price A and price B in the historical used car transaction data; the output is configured with weight coefficients a and b. In order to optimize the final price prediction C, it is necessary to select the optimal weight coefficients a and b, which can be optimized by grid search method or gradient descent method.
[0150] Optionally, performing weighted fusion on the first value prediction model and the second value prediction model to obtain a third value prediction model includes:
[0151] According to the first value prediction model, obtaining a first prediction result of the first value prediction model;
[0152] According to the second value prediction model, obtaining a second prediction result of the second value prediction model;
[0153] A third value prediction model is obtained by weighted fusion of the first prediction result and the second prediction result. The third value prediction model is configured as follows:
[0154] The third prediction result = a*first prediction result + b*second prediction result;
[0155] Among them, the input of the third value prediction model is the first prediction result and the second prediction result, the output of the third value prediction model is the third prediction result, a is the weight coefficient of the first value prediction model, and b is the weight coefficient of the second value prediction model.
[0156] Optionally, weighted parameter optimization is performed on the third value prediction model based on the preprocessed historical used car transaction data to obtain a used car value prediction model, including:
[0157] According to the pre-processed historical used car transaction data and the third value prediction model, a first set of weight coefficients is obtained based on the grid search method;
[0158] According to the pre-processed historical used car transaction data and the third value prediction model, a second set of weight coefficients is obtained based on the gradient descent method;
[0159] Based on the first set of weight coefficients and / or the second set of weight coefficients, weighted parameters of the third value prediction model are optimized to obtain a used car value prediction model.
[0160] Optionally, also include:
[0161] Establishing a value correction model, the value correction model is configured to take the model and the associated events of the model as input and take the price change as output;
[0162] According to the pre-processed historical used car transaction data, the weighted parameters of the third value prediction model are optimized to obtain a used car value prediction model, including:
[0163] According to the pre-processed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain an optimized third value prediction model;
[0164] According to the optimized third value prediction model and value correction model, a used car value prediction model is obtained.
[0165] Considering that the value of used cars is not only affected by the condition of the vehicle itself, but also by market changes, for example, after a new model is released, the value of used cars of the same model will generally drop significantly, and for existing AI models,
[0166] Optionally, establish a value revision model, including:
[0167] Establishing an initial value correction model, which is configured as a regression analysis model, taking vehicle models and vehicle model related events as inputs and taking price changes as outputs;
[0168] Based on historical second-hand car transaction data, obtain the transaction price distribution of the same second-hand cars with similar vehicle characteristics;
[0169] Based on the price distribution, obtain at least one price revision data;
[0170] According to the price correction data, matching the associated events corresponding to the price correction data;
[0171] The regression analysis model is trained according to the price correction data, the associated events corresponding to the price correction data, and the vehicle model corresponding to the price correction data to obtain a value correction model.
[0172] By adopting the above method, by analyzing the transaction price distribution of the same used cars with similar vehicle characteristics, the price change trend of the used car can be obtained. According to the price change trend of the used car, at least one price correction data can be obtained. By analyzing the reasons leading to the price correction, that is, matching the associated events corresponding to the price correction data, the impact on the value of the used car at different times can be obtained to obtain a value correction model.
[0173] Optionally, based on historical used car transaction data, the transaction price distribution of the same used cars with similar vehicle characteristics is obtained, including:
[0174] According to the historical second-hand car transaction data, the data is classified according to the car model, and the historical second-hand car transaction data of at least one car model is obtained;
[0175] Clustering the historical second-hand car transaction data of each model according to vehicle characteristics to obtain the first clustering result;
[0176] According to the first clustering result, the transaction price and transaction time of each transaction in each cluster are obtained;
[0177] Clustering the data of each cluster in the first clustering result again according to the transaction price and transaction time to obtain a second clustering result;
[0178] According to the second clustering result, the transaction price distribution of the same used cars with similar characteristics is obtained.
[0179] By adopting the above method, through two clustering, we can obtain the clustering results according to time and vehicle characteristics. Based on this result, we can analyze that at a certain time, the used cars of a certain model have a unified price fluctuation. We can obtain events or public opinions at a time close to this time as related events, which is convenient for establishing a value correction model.
[0180] Optionally, a method for obtaining a used car value prediction model based on the optimized third value prediction model and the value correction model includes:
[0181] According to the third value prediction model, obtaining a model update time of the third value prediction model;
[0182] According to the model update time, obtain the associated events of each model after the model update time;
[0183] Obtain value revision data based on the associated events and value revision models of each vehicle model;
[0184] A used car value prediction model is obtained based on the third value prediction model and the value correction data.
[0185] Specifically, a pseudo code example to implement the above method:
[0186]
[0187]
[0188]
[0189] S6. Input the information of the vehicle to be evaluated into a used car value prediction model to obtain the used car price of the vehicle to be evaluated.
[0190] Optionally, the information of the vehicle to be evaluated is input into a used car value prediction model to obtain the used car value of the vehicle to be evaluated, and points are added or evaluated based on the used car value to obtain the used car price of the vehicle to be evaluated.
[0191] Optionally, the added points may be a preset proportion or a preset fixed value. After the value of the used car is analyzed, the used car price of the vehicle to be evaluated can be quickly evaluated based on the actual business situation.
[0192] This solution introduces an indicator system generated by the experience of traditional second-hand car traders, integrates the traditional second-hand car value analysis method with the AI second-hand car value analysis method, and improves the prediction accuracy, interpretability and robustness of the entire prediction method. At least it solves the problem that the existing second-hand car price prediction model generally only relies on the characteristics of the vehicle and historical transaction data for prediction, with low accuracy, low interpretability of the prediction model, and poor robustness, and provides a second-hand car price prediction method with high accuracy, high interpretability and strong robustness.
[0193] Example 3
[0194] A used car price prediction system based on multi-model fusion, including a transaction data analysis platform and a transaction data storage platform;
[0195] The transaction data analysis platform is configured as follows:
[0196] Obtain historical second-hand car transaction data and pre-process the historical second-hand car transaction data;
[0197] According to the pre-processed historical used car transaction data, an MLP model is established and trained to obtain the first value prediction model;
[0198] According to the pre-processed historical second-hand car transaction data, the second value prediction model is obtained based on the indicator system method;
[0199] Performing weighted fusion on the first value prediction model and the second value prediction model to obtain a third value prediction model;
[0200] According to the pre-processed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain a second-hand car value prediction model;
[0201] Inputting the information of the vehicle to be evaluated into the used car value prediction model to obtain the used car price of the vehicle to be evaluated;
[0202] The transaction data storage platform is configured as follows:
[0203] The historical second-hand car transaction data is stored, and in response to a request from the transaction data analysis platform, the historical second-hand car transaction data is sent to the transaction data analysis platform.
[0204] Optionally, based on the preprocessed historical used car transaction data, an MLP model is established and trained to obtain a first value prediction model, including:
[0205] According to the preprocessed historical used car transaction data, an MLP model is established, and the MLP model is configured to include an input layer, a hidden layer, and an output layer;
[0206] According to the preprocessed historical used car transaction data, a training set and a test set are obtained;
[0207] The MLP model is trained according to the training set and the test set to obtain the first value prediction model.
[0208] Optionally, according to the pre-processed historical used car transaction data, based on the indicator system method, a second value prediction model is obtained, including:
[0209] According to the pre-processed historical second-hand car transaction data, key indicators in the historical second-hand car transaction data are obtained;
[0210] According to the key indicators in the historical second-hand car transaction data, based on the indicator system method, the second-hand car value indicator system is obtained;
[0211] According to the second-hand car value index system, the second-hand car value prediction rules are obtained;
[0212] According to the used car value prediction rules, a second value prediction model is obtained.
[0213] Optionally, performing weighted fusion on the first value prediction model and the second value prediction model to obtain a third value prediction model includes:
[0214] According to the first value prediction model, obtaining a first prediction result of the first value prediction model;
[0215] According to the second value prediction model, obtaining a second prediction result of the second value prediction model;
[0216] A third value prediction model is obtained by weighted fusion of the first prediction result and the second prediction result. The third value prediction model is configured as follows:
[0217] The third prediction result = a*first prediction result + b*second prediction result;
[0218] Among them, the input of the third value prediction model is the first prediction result and the second prediction result, the output of the third value prediction model is the third prediction result, a is the weight coefficient of the first value prediction model, and b is the weight coefficient of the second value prediction model.
[0219] Optionally, weighted parameter optimization is performed on the third value prediction model based on the preprocessed historical used car transaction data to obtain a used car value prediction model, including:
[0220] According to the pre-processed historical used car transaction data and the third value prediction model, a first set of weight coefficients is obtained based on the grid search method;
[0221] According to the pre-processed historical used car transaction data and the third value prediction model, a second set of weight coefficients is obtained based on the gradient descent method;
[0222] Based on the first set of weight coefficients and / or the second set of weight coefficients, weighted parameters of the third value prediction model are optimized to obtain a used car value prediction model.
[0223] Optionally, also include:
[0224] Establishing a value correction model, the value correction model is configured to take the model and the associated events of the model as input and take the price change as output;
[0225] According to the pre-processed historical used car transaction data, the weighted parameters of the third value prediction model are optimized to obtain a used car value prediction model, including:
[0226] According to the pre-processed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain an optimized third value prediction model;
[0227] According to the optimized third value prediction model and value correction model, a used car value prediction model is obtained.
[0228] Optionally, establish a value revision model, including:
[0229] Establishing an initial value correction model, which is configured as a regression analysis model, taking vehicle models and vehicle model related events as inputs and taking price changes as outputs;
[0230] Based on historical second-hand car transaction data, obtain the transaction price distribution of the same second-hand cars with similar vehicle characteristics;
[0231] Based on the price distribution, obtain at least one price revision data;
[0232] According to the price correction data, matching the associated events corresponding to the price correction data;
[0233] The regression analysis model is trained according to the price correction data, the associated events corresponding to the price correction data, and the vehicle model corresponding to the price correction data to obtain a value correction model.
[0234] Optionally, based on historical used car transaction data, the transaction price distribution of the same used cars with similar vehicle characteristics is obtained, including:
[0235] According to the historical second-hand car transaction data, the data is classified according to the car model, and the historical second-hand car transaction data of at least one car model is obtained;
[0236] Clustering the historical second-hand car transaction data of each model according to vehicle characteristics to obtain the first clustering result;
[0237] According to the first clustering result, the transaction price and transaction time of each transaction in each cluster are obtained;
[0238] Clustering the data of each cluster in the first clustering result again according to the transaction price and transaction time to obtain a second clustering result;
[0239] According to the second clustering result, the transaction price distribution of the same used cars with similar characteristics is obtained.
[0240] Example 4
[0241] This embodiment provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement any of the above methods.
[0242] Specifically, Figure 2 As shown, Figure 2This is a structural diagram of a computer device in the present application, and the computer device may include: a processor 101, such as a central processing unit (CPU), a communication bus 102, a user interface 104, a network interface 103, and a memory 105. Among them, the communication bus 102 is used to realize the connection and communication between these components. The user interface 104 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 104 may also include a standard wired interface and a wireless interface. The network interface 103 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 105 may optionally be a storage device independent of the aforementioned processor 101. The memory 105 may be a high-speed random access memory (RAM) memory, or it may be a stable non-volatile memory (NVM), such as at least one disk storage; the processor 101 may be a general-purpose processor, including a central processing unit, a network processor, etc., or it may be a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component.
[0243] Those skilled in the art will appreciate that Figure 2 The structure shown in the figure does not constitute a limitation on the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.
[0244] like Figure 2 As shown, the memory 105 as a storage medium may include an operating system, a network communication module, a user interface module, and an application program for implementing a used car price prediction method based on multi-model fusion.
[0245] exist Figure 2 In the electronic device shown, the network interface 103 is mainly used for data communication with the network server; the user interface 104 is mainly used for data interaction with the user; the processor 101 and the memory 105 in the present application can be set in the electronic device, and the electronic device calls the application stored in the memory 105 for implementing a used car price prediction method based on multi-model fusion through the processor 101 to implement the above method.
[0246] Example 5
[0247] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and a processor executes the computer program to implement any of the above methods.
[0248] In some embodiments, the computer readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or various devices including one or any combination of the above memories. The computer may be various computing devices including intelligent terminals and servers.
[0249] In the above embodiments of the present disclosure, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0250] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.
[0251] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0252] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0253] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a non-volatile storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present disclosure. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.
[0254] The above are only preferred embodiments of the present disclosure. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as the protection scope of the present disclosure.
Claims
1. A method for predicting second-hand car prices based on multi-model fusion, characterized in that: include: Acquire historical second-hand car transaction data, and pre-process the historical second-hand car transaction data; According to the pre-processed historical used car transaction data, an MLP model is established and trained to obtain the first value prediction model; According to the pre-processed historical second-hand car transaction data, the second value prediction model is obtained based on the indicator system method; Performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model; According to the preprocessed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain a second-hand car value prediction model; The information of the vehicle to be evaluated is input into the second-hand car value prediction model to obtain the second-hand car price of the vehicle to be evaluated.
2. The method for predicting second-hand car prices based on multi-model fusion according to claim 1, characterized in that: The method of establishing an MLP model and training the MLP model based on the preprocessed historical used car transaction data to obtain a first value prediction model includes: Establishing an MLP model based on the preprocessed historical used car transaction data, wherein the MLP model is configured to include an input layer, a hidden layer, and an output layer; According to the preprocessed historical used car transaction data, a training set and a test set are obtained; The MLP model is trained according to the training set and the test set to obtain a first value prediction model.
3. The method for predicting second-hand car prices based on multi-model fusion according to claim 1, characterized in that: The second value prediction model is obtained based on the pre-processed historical used car transaction data and the indicator system method, including: The method of obtaining key indicators in the historical second-hand car transaction data based on the pre-processed historical second-hand car transaction data; According to the key indicators in the historical used car transaction data, based on the indicator system method, a used car value indicator system is obtained; According to the second-hand car value index system, a second-hand car value prediction rule is obtained; According to the second-hand car value prediction rule, a second value prediction model is obtained.
4. The method for predicting second-hand car prices based on multi-model fusion according to claim 1, characterized in that: The step of performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model includes: According to the first value prediction model, obtaining a first prediction result of the first value prediction model; According to the second value prediction model, obtaining a second prediction result of the second value prediction model; A third value prediction model is obtained by weighted fusion of the first prediction result and the second prediction result. The third value prediction model is configured as follows: The third prediction result = a*first prediction result + b*second prediction result; Among them, the input of the third value prediction model is the first prediction result and the second prediction result, the output of the third value prediction model is the third prediction result, a is the weight coefficient of the first value prediction model, and b is the weight coefficient of the second value prediction model.
5. The method for predicting second-hand car prices based on multi-model fusion according to claim 1, characterized in that: The method of optimizing the weighted parameters of the third value prediction model based on the preprocessed historical second-hand car transaction data to obtain the second-hand car value prediction model includes: According to the pre-processed historical used car transaction data and the third value prediction model, a first set of weight coefficients is obtained based on the grid search method; According to the pre-processed historical used car transaction data and the third value prediction model, a second set of weight coefficients is obtained based on the gradient descent method; Based on the first set of weight coefficients and / or the second set of weight coefficients, weighted parameters of the third value prediction model are optimized to obtain a used car value prediction model.
6. The method for predicting second-hand car prices based on multi-model fusion according to claim 1, characterized in that: Also includes: Establishing a value correction model, wherein the value correction model is configured to take a vehicle model and an event associated with the vehicle model as input and take a price change as output; The method of optimizing the weighted parameters of the third value prediction model based on the preprocessed historical second-hand car transaction data to obtain the second-hand car value prediction model includes: The weighted parameter optimization of the third value prediction model is performed according to the preprocessed historical second-hand car transaction data to obtain an optimized third value prediction model; A used car value prediction model is obtained based on the optimized third value prediction model and the value correction model.
7. The method for predicting second-hand car prices based on multi-model fusion according to claim 6, characterized in that: The establishing of the value correction model comprises: Establishing an initial value correction model, wherein the initial value correction model is configured as a regression analysis model, with the vehicle model and the associated events of the vehicle model as inputs and the price change as output; Based on historical second-hand car transaction data, obtain the transaction price distribution of the same second-hand cars with similar vehicle characteristics; According to the price distribution, obtaining at least one price revision data; According to the price correction data, matching the associated events corresponding to the price correction data; The regression analysis model is trained according to the price correction data, the associated events corresponding to the price correction data, and the vehicle model corresponding to the price correction data to obtain a value correction model.
8. The method for predicting second-hand car prices based on multi-model fusion according to claim 7, characterized in that: The method of obtaining the transaction price distribution of the same used cars with similar vehicle characteristics based on the historical used car transaction data includes: According to the historical second-hand car transaction data, the data is classified according to the car model, and the historical second-hand car transaction data of at least one car model is obtained; Clustering the historical second-hand car transaction data of each model according to vehicle characteristics to obtain the first clustering result; According to the first clustering result, the transaction price and transaction time of each transaction in each cluster are obtained; Clustering the data of each cluster in the first clustering result again according to the transaction price and transaction time to obtain a second clustering result; According to the second clustering result, the transaction price distribution of the same used cars with similar characteristics is obtained.
9. A used car price prediction system based on multi-model fusion, characterized in that: Including transaction data analysis platform and transaction data storage platform; The transaction data analysis platform is configured to: Acquire historical second-hand car transaction data, and pre-process the historical second-hand car transaction data; According to the pre-processed historical used car transaction data, an MLP model is established and trained to obtain the first value prediction model; According to the pre-processed historical second-hand car transaction data, the second value prediction model is obtained based on the indicator system method; Performing weighted fusion according to the first value prediction model and the second value prediction model to obtain a third value prediction model; According to the preprocessed historical second-hand car transaction data, weighted parameters of the third value prediction model are optimized to obtain a second-hand car value prediction model; Inputting the information of the vehicle to be evaluated into the second-hand car value prediction model to obtain the second-hand car price of the vehicle to be evaluated; The transaction data storage platform is configured as follows: The historical second-hand car transaction data is stored, and in response to a request from the transaction data analysis platform, the historical second-hand car transaction data is sent to the transaction data analysis platform.
10. A device, characterized in that: The device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method according to any one of claims 1 to 8.