Method and system for predicting travel route selection behavior based on survey and data analysis

Abstract

The invention discloses a method for predicting travel route selection behaviors based on investigation and data analysis, and the method comprises the steps: collecting data, including historical traffic data, real-time traffic data, user behavior data and environmental factors; preprocessing the data, including data cleaning, feature extraction and standardization, to obtain preprocessed data; constructing a structural equation model by taking the preprocessed data as input and the predicted route probability and travel time as output; training the structural equation model based on the fitting degree between the index evaluation and the covariance matrix; and predicting travel route selection behaviors based on the trained structural equation model. According to the method, the individual travel experience and the wider traffic system efficiency are improved by providing prediction insights for optimizing travel behaviors, reducing traffic congestion and reducing environmental influence to the maximum extent.

Description

Technical Field

[0001] The present invention belongs to the field of traffic planning and data analysis, and in particular to a method and system for predicting travel route selection behavior based on investigation and data analysis. Background Art

[0002] The fundamental cause of traffic congestion is the imbalance between supply and demand, that is, the contradiction caused by the mismatch between the travel needs of travelers and the road traffic supply resources. Existing research on solving the supply and demand balance problem mainly starts from two aspects: reducing demand and increasing supply, namely traffic demand management and traffic supply management. With the rapid expansion of urbanization and road networks, the ability to predict route selection behavior is crucial for traffic planning, traffic management, and environmental impact assessment. Existing prediction systems are usually limited to historical data and simple models, and lack adaptability to real-time changes in traffic conditions, road availability, and user preferences.

[0003] Furthermore, traditional approaches do not take into account the complex interactions between individual traveler behavior, socioeconomic factors, and road network dynamics. With advances in GPS, mobile devices, and intelligent transportation system data collection technology, an advanced forecasting system is needed that combines these data sources with modern machine learning techniques to provide accurate and dynamic predictions of travel behavior. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method and system for predicting travel route selection behavior based on surveys and data analysis. Used to accurately predict vehicle mileage and people's travel route selection behavior. The system aims to meet the growing demand for more efficient and adaptive traffic management by leveraging real-time data, historical traffic information and machine learning algorithms to optimize traffic flow, reduce travel time and improve decision-making for individual travelers and transportation departments. The invention aims to improve individual travel experiences and the efficiency of the broader transportation system by providing predictive insights that optimize travel behavior, reduce traffic congestion and minimize environmental impact.

[0005] Technical solution: A method for predicting travel route selection behavior based on survey and data analysis of the present invention comprises the following steps:

[0006] Step 1: Collect data, including historical traffic data, real-time traffic data, user behavior data and environmental factors;

[0007] Step 2: preprocess the data, including data cleaning, feature extraction and standardization, to obtain preprocessed data;

[0008] Step 3: Using the preprocessed data as input and the predicted route probability and travel time as output, a structural equation model is constructed;

[0009] Step 4: Based on the goodness of fit between the indicator evaluation and the covariance matrix, the structural equation model is trained;

[0010] Step 5: Based on the trained structural equation model, travel route choice behavior is predicted.

[0011] Furthermore, step 1 is specifically as follows: Collect historical traffic data: collected from GPS trackers, road sensors and traffic cameras, providing timestamps, speed and traffic volume; Collect real-time traffic data: including congestion levels, accidents, weather conditions and road closures from IoT devices and APIs; Collect user behavior data: obtained through travel behavior surveys, including route preferences and decision-making patterns; Collect environmental factors: weather conditions, traffic flow, etc. t , oil price P f , event E t ; Data format: Dt = (T, L, V, W t , P f , E t ), where T is the timestamp of the travel data, L = (x, y) the longitude and latitude of different points on the road, and V is the average travel speed.

[0012] Furthermore, step 2 specifically includes the following steps:

[0013] Step 2.1. Data preparation: Data cleaning: Use statistical thresholds to remove outliers:

[0014] Ot={x i ∣x i -μ x ∣>k·σ x}

[0015] Among them, x i is a data point, μ x is the mean value, σ x is the standard deviation, k is the scaling factor;

[0016] Step 2.2, feature extraction: calculate derived features, including congestion time CI and travel time T t , CI = number of vehicles per kilometer / capacity, T t = distance / average speed;

[0017] Step 2.3, Normalization: Normalize the input features to the range [0,1].

[0018] Furthermore, step 3 specifically includes the following steps:

[0019] Step 3.1, model establishment: Input variables: real-time travel characteristics R = [T, L, CI, W t , P f , Et ]; Historical mode: H t =[T t-1 ,T t-2 ,…,T t-n ]; Output variable: predicted route probability: P0(route i |R,H t ); Travel time: T t =f(R,H t );

[0020] Step 3.2, model building: constructing a structural equation model including exogenous indicators and endogenous indicators;

[0021] Exogenous indicators:

[0022] x=Λxξ+δ

[0023] Among them, x is a vector composed of exogenous indicators; ξ is an exogenous latent variable; Λx is the coefficient matrix of x to ξ, that is, the factor loading in factor analysis; δ represents the measurement error of x;

[0024] Endogenous indicators:

[0025] y=Λyη+ε

[0026] Among them, y is a vector composed of endogenous indicators; η is an endogenous latent variable; Λy is the coefficient matrix of y for η; ε represents the measurement error of y,

[0027] Endogenous latent variable expression:

[0028] η=Bη+Γξ+ζ

[0029] η is the endogenous latent variable; B is the coefficient matrix between the endogenous latent variables, indicating the mutual influence between the endogenous latent variables; Γ is the coefficient matrix of the exogenous latent variable on the endogenous latent variable; ξ is the exogenous latent variable; ζ represents the error term of the structural equation.

[0030] Furthermore, step 4 specifically includes the following steps:

[0031] Step 4.1, the model implicit covariance matrix is ​​as follows:

[0032] Σ=Λy(IB) -1 ΓΦΓ T (IB) -T Λy T +Λy(IB) -1 Ψζ(IB) -T Λy T +Θ∈

[0033] Where Λy is the coefficient matrix of y to η, I is the identity matrix, B is the coefficient matrix between endogenous latent variables, Γ is the regression coefficient between exogenous latent variables and endogenous latent variables, Φ, Ψ, Θ∈ are the covariances of the errors between exogenous latent variables, between endogenous latent variable errors, and between observed variables, ζ represents the error term of the structural equation, and T represents the transpose of the matrix;

[0034] Maximum Likelihood Estimation:

[0035] FML=log|Σ|+tr(SΣ -1 )-log|S|-p

[0036] In the formula, Σ represents a covariance matrix, |Σ| represents the determinant of Σ, log|Σ| is the logarithm of the determinant, S represents the sample covariance matrix, Σ -1 is the inverse matrix of Σ, tr(SΣ -1 ) is S and Σ -1 The trace of the product, p is the number of observed variables;

[0037] Weighted least squares function:

[0038] F WLS =(s-σ) T W(s-σ)

[0039] Where s represents the vectorized elements of the sample covariance matrix; σ represents the vectorized elements of the model implicit covariance matrix; W represents the weighting matrix;

[0040] Iteratively update the parameters according to the gradient of the objective function:

[0041]

[0042] In the formula, θ t is the parameter of the tth iteration, η is the learning rate, and F is the gradient;

[0043] Use indicators to assess the goodness of fit between S and Σ; if the fit is poor, modify the model and re-estimate; output estimation parameters include factor loading Λ, path coefficients (B, Γ), covariance (Φ, Ψ) and error variance Θ; fit index is an indicator of how well the model fits the data; residual matrix: the difference between S and Σ, highlighting areas of model improvement;

[0044] Step 4.2: Continue learning and feedback integration, use actual travel data for continuous training, and dynamically adjust predictions.

[0045] Furthermore, the modifying the model includes adding paths and adjusting constraints.

[0046] The present invention also discloses a system for predicting travel route selection behavior based on survey and data analysis, including a data collection module, a data analysis module, a behavior analysis module, a route selection optimization and recommendation module, and a machine learning model module;

[0047] The data collection module integrates data from multiple sources;

[0048] The data analysis module prediction engine is driven by a machine learning algorithm and uses the collected data to predict average mileage and route selection behavior;

[0049] The behavior analysis module analyzes factors that affect travel behavior;

[0050] The route selection optimization and recommendation module provides optimized route recommendations to users based on predicted average mileage, estimated travel time, traffic congestion, and user preferences;

[0051] The machine learning model module is trained on a large amount of travel history and traffic pattern datasets.

[0052] The present invention also discloses a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method of the present invention.

[0053] The present invention also discloses a computer-readable storage medium on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the steps of the method of the present invention are implemented.

[0054] The present invention also discloses a computer program product, comprising a computer program / instruction, which implements the steps of the method of the present invention when executed by a processor.

[0055] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0056] (1) The present invention achieves higher accuracy in predicting travel route selection by using advanced machine learning models to continuously learn from real-time data. The system is able to provide real-time optimized route suggestions, thereby reducing individual user travel time and reducing overall road congestion, especially during peak hours. By optimizing driving routes based on predicted traffic patterns, the invention reduces unnecessary detours, stop-and-go stops, and idling, thereby improving fuel efficiency and reducing carbon dioxide emissions.

[0057] (2) The invention addresses the growing need for more efficient and adaptive traffic management by leveraging real-time data, historical traffic information, and machine learning algorithms to optimize traffic flows, reduce travel times, and improve decision making for individual travelers and transportation agencies. The invention is intended to improve both individual travel experiences and the efficiency of the broader transportation system by providing predictive insights that optimize travel behavior, reduce traffic congestion, and minimize environmental impact.

[0058] (3) The present invention can combine directly observed variables (e.g., travel time, road conditions) with latent variables (e.g., traveler preferences, perceived convenience). This allows the inclusion of unmeasured but theoretically important factors, such as personal attitudes, socioeconomic status, or perceived route utility. It supports simultaneous modeling of direct, indirect, and mediating effects, allowing for a more comprehensive understanding of how various factors (e.g., congestion, weather, cost) affect route choice. Unlike simple regression models, the model can capture feedback loops and correlations between dependent and independent variables. The model can accommodate multiple types of relationships: such as causal paths and measurement models for inferring latent variables from observed data. This modularity enables the model to adapt to different use cases, such as urban vs. rural route prediction. The introduction of latent variables such as perceived convenience or environmental awareness provides a new way to quantify behavioral factors that are difficult to measure directly. Combining discrete choice theory (e.g., logistic models) with the latent variable approach of structural equation models creates a hybrid model that connects behavioral economics and statistical modeling. This improves the understanding of the trade-offs that travelers make between time, cost, and convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of the structural equation model of the present invention.

[0060] Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0061] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0062] like Figure 1 , Figure 2 As shown, a method for predicting travel route selection behavior based on investigation and data analysis of the present invention comprises the following steps:

[0063] Step 1: Collect data, including historical traffic data, real-time traffic data, user behavior data and environmental factors;

[0064] Step 2: preprocess the data, including data cleaning, feature extraction and standardization, to obtain preprocessed data;

[0065] Step 3: Using the preprocessed data as input and the predicted route probability and travel time as output, a structural equation model is constructed;

[0066] Step 4: Based on the goodness of fit between the indicator evaluation and the covariance matrix, the structural equation model is trained;

[0067] Step 5: Based on the trained structural equation model, travel route choice behavior is predicted.

[0068] Furthermore, step 1 is specifically as follows: Collect historical traffic data: collected from GPS trackers, road sensors and traffic cameras, providing timestamps, speed and traffic volume; Collect real-time traffic data: including congestion levels, accidents, weather conditions and road closures from IoT devices and APIs; Collect user behavior data: obtained through travel behavior surveys, including route preferences and decision-making patterns; Collect environmental factors: weather conditions, traffic flow, etc. t , oil price P f , event E t ; Data format: Dt = (T, L, V, W t , P f , E t ), where T is the timestamp of the travel data, L = (x, y) the longitude and latitude of different points on the road, and V is the average travel speed.

[0069] Furthermore, step 2 specifically includes the following steps:

[0070] Step 2.1. Data preparation: Data cleaning: Use statistical thresholds to remove outliers:

[0071] Ot={x i ∣x i -μ x ∣>k·σ x}

[0072] Among them, x i is a data point, μ x is the mean value, σ x is the standard deviation, k is the scaling factor;

[0073] Step 2.2, feature extraction: calculate derived features, including congestion time CI and travel time T t , CI = number of vehicles per kilometer / capacity, T t = distance / average speed;

[0074] Step 2.3, Normalization: Normalize the input features to the range [0,1].

[0075] Furthermore, step 3 specifically includes the following steps:

[0076] Step 3.1, model establishment: Input variables: real-time travel characteristics R = [T, L, CI, W t , P f , E t ]; History mode: H t =[T t-1 ,T t-2 ,…,T t-n ]; Output variable: predicted route probability: P0(route i |R,H t ); Travel time: T t =f(R,H t );

[0077] Step 3.2, model building: constructing a structural equation model including exogenous indicators and endogenous indicators;

[0078] Exogenous indicators:

[0079] x=Λxξ+δ

[0080] Among them, x is a vector composed of exogenous indicators; ξ is an exogenous latent variable; Λx is the coefficient matrix of x to ξ, that is, the factor loading in factor analysis; δ represents the measurement error of x;

[0081] Endogenous indicators:

[0082] y=Λyη+ε

[0083] Among them, y is a vector composed of endogenous indicators; η is an endogenous latent variable; Λy is the coefficient matrix of y for η; ε represents the measurement error of y,

[0084] Endogenous latent variable expression:

[0085] η=Bη+Γξ+ζ

[0086] η is the endogenous latent variable; B is the coefficient matrix between the endogenous latent variables, indicating the mutual influence between the endogenous latent variables; Γ is the coefficient matrix of the exogenous latent variable on the endogenous latent variable; ξ is the exogenous latent variable; ζ represents the error term of the structural equation.

[0087] Furthermore, step 4 specifically includes the following steps:

[0088] Step 4.1, the model implicit covariance matrix is ​​as follows:

[0089] Σ=Λy(IB) -1 ΓΦΓ T (IB) -T Λy T +Λy(IB) -1 Ψζ(IB) -T Λy T +Θ∈

[0090] Where Λy is the coefficient matrix of y to η, I is the identity matrix, B is the coefficient matrix between endogenous latent variables, Γ is the regression coefficient between exogenous latent variables and endogenous latent variables, Φ, Ψ, Θ∈ are the covariances of the errors between exogenous latent variables, between endogenous latent variable errors, and between observed variables, ζ represents the error term of the structural equation, and T represents the transpose of the matrix;

[0091] Maximum Likelihood Estimation:

[0092] FML=log|Σ|+tr(SΣ -1 )-log|S|-p

[0093] In the formula, Σ represents a covariance matrix, |Σ| represents the determinant of Σ, log|Σ| is the logarithm of the determinant, S represents the sample covariance matrix, Σ -1 is the inverse matrix of Σ, tr(SΣ -1 ) is S and Σ -1 The trace of the product, p is the number of observed variables;

[0094] Weighted least squares function:

[0095] F WLS =(s-σ) T W(s-σ)

[0096] Where s represents the vectorized elements of the sample covariance matrix; σ represents the vectorized elements of the model implicit covariance matrix; W represents the weighting matrix;

[0097] Iteratively update the parameters according to the gradient of the objective function:

[0098]

[0099] In the formula, θ t is the parameter of the tth iteration, η is the learning rate, and F is the gradient;

[0100] Use indicators to assess the goodness of fit between S and Σ; if the fit is poor, modify the model and re-estimate; output estimation parameters include factor loading Λ, path coefficients (B, Γ), covariance (Φ, Ψ) and error variance Θ; fit index is an indicator of how well the model fits the data; residual matrix: the difference between S and Σ, highlighting areas of model improvement;

[0101] Step 4.2: Continue learning and feedback integration, use actual travel data for continuous training, and dynamically adjust predictions.

[0102] Furthermore, the modifying the model includes adding paths and adjusting constraints.

[0103] The system of the present invention comprises the following:

[0104] 1. Data collection; this module integrates data from multiple sources, including:

[0105] GPS data: collected from vehicles and mobile devices to monitor travel routes, speed, and vehicle location.

[0106] Traffic sensor data: Includes data from road sensors, traffic cameras, and vehicle detectors to provide real-time traffic conditions.

[0107] Historical traffic data: collected from traffic management systems and other databases to analyze long-term traffic trends.

[0108] User Data: includes personal preferences, socioeconomic factors, and demographic information collected from surveys or user profiles.

[0109] Environmental data: such as weather conditions and roadwork updates.

[0110] 2. Data analysis: The prediction engine is driven by machine learning algorithms and uses the collected data to predict average mileage and route selection behavior. Key features include:

[0111] Average Mileage Prediction: A regression model trained on historical average mileage data and real-time traffic inputs can predict the total number of kilometers a vehicle will travel within a given time frame or on a specific route.

[0112] Route Choice Behavior Prediction: A classification model that predicts the likelihood that a user will choose a specific route based on real-time and historical data (e.g., traffic congestion, travel time, tolls, and personal preferences).

[0113] Data cleaning: Remove noise, handle missing data, and filter out irrelevant information (such as outliers or corrupted data).

[0114] Data normalization: Normalize data from different sources (e.g., GPS coordinates, traffic sensor inputs) into a consistent format and units for easier analysis.

[0115] Feature Engineering: Extracting relevant features from the data, such as travel time, distance, traffic congestion, weather conditions, and socioeconomic factors that may influence route choice.

[0116] 3. Behavior analysis: This module analyzes factors that affect travel behavior, including:

[0117] Socioeconomic variables: income level, occupation, vehicle ownership, and other demographic factors that may influence route choice behavior.

[0118] Personal preferences: For example, some users may prioritize travel time over cost, while others may prefer routes that avoid highways.

[0119] Past behavior: Analyze user travel history to detect patterns and preferences that can inform future predictions.

[0120] The behavior analysis module combines real-time traffic data to analyze various factors to fully understand an individual's possible travel behavior.

[0121] 4. Route selection optimization and recommendation: The system provides users with optimized route recommendations based on predicted average mileage, estimated travel time, traffic congestion, and user preferences. The route selection optimization function calculates multiple route alternatives and provides the most efficient option based on the following factors:

[0122] Real-time traffic conditions (e.g., incidents, road closures, construction).

[0123] Environmental impact (e.g. fuel efficiency, emissions).

[0124] User-specific preferences (e.g. shortest distance, lowest toll).

[0125] 5. Machine learning models: Machine learning models are trained on a large dataset of travel history and traffic patterns. The system combines various types of models, such as:

[0126] Regression models: Used to predict continuous variables, such as the number of kilometers a vehicle is likely to travel.

[0127] Classification models: used to classify travelers’ preferences and predict route choices.

[0128] Clustering technology: Segment users based on their travel behavior and optimize predictions for different user groups.

[0129] System adaptability:

[0130] The system is designed to adapt to a variety of geographic areas and can be customized based on available data and specific local traffic conditions. The machine learning model is continuously updated with new data, ensuring that predictions remain accurate as traffic patterns evolve.

Claims

1. A method for predicting travel route selection behavior based on survey and data analysis, characterized in that: The steps include: Step 1: Collect data, including historical traffic data, real-time traffic data, user behavior data and environmental factors; Step 2: preprocess the data, including data cleaning, feature extraction and standardization, to obtain preprocessed data; Step 3: Using the preprocessed data as input and the predicted route probability and travel time as output, a structural equation model is constructed; Step 4: Based on the goodness of fit between the indicator evaluation and the covariance matrix, the structural equation model is trained; Step 5: Based on the trained structural equation model, travel route choice behavior is predicted.

2. A method for predicting travel route selection behavior based on survey and data analysis according to claim 1, characterized in that: Step 1 is as follows: Collect historical traffic data: collected from GPS trackers, road sensors and traffic cameras, providing timestamps, speed and traffic volume; Collect real-time traffic data: including congestion levels, accidents, weather conditions and road closures from IoT devices and APIs; Collect user behavior data: obtained through travel behavior surveys, including route preferences and decision-making patterns; Collect environmental factors: weather conditions, traffic flow, etc. t , oil price P f , event E t ; Data format: Dt = (T, L, V, W t , P f , E t ), where T is the timestamp of the travel data, L = (x, y) the longitude and latitude of different points on the road, and V is the average travel speed.

3. The method for predicting travel route selection behavior based on survey and data analysis according to claim 1, characterized in that: Step 2 specifically includes the following steps: Step 2.

1. Data preparation: Data cleaning: Use statistical thresholds to remove outliers: Ot={x i ∣x i -μ x ∣>k·σ x } Among them, x i is a data point, μ x is the mean value, σ x is the standard deviation, k is the scaling factor; Step 2.2, feature extraction: calculate derived features, including congestion time CI and travel time T t , CI = number of vehicles per kilometer / capacity, T t = distance / average speed; Step 2.3, Normalization: Normalize the input features to the range [0,1].

4. The method for predicting travel route selection behavior based on survey and data analysis according to claim 1, characterized in that: Step 3 specifically includes the following steps: Step 3.1, model establishment: Input variables: real-time travel characteristics R = [T, L, CI, W t , P f , E t ]; History mode: H t =[T t-1 ,T t-2 ,…,T t-n ]; Output variable: predicted route probability: P0(route i |R,H t ); Travel time: T t =f(R,H t ); Step 3.2, model building: constructing a structural equation model including exogenous indicators and endogenous indicators; Exogenous indicators: x=Λxξ+δ Among them, x is a vector composed of exogenous indicators; ξ is an exogenous latent variable; Λx is the coefficient matrix of x to ξ, that is, the factor loading in factor analysis; δ represents the measurement error of x; Endogenous indicators: y=Λyη+ε Among them, y is a vector composed of endogenous indicators; η is an endogenous latent variable; Λy is the coefficient matrix of y for η; ε represents the measurement error of y, Endogenous latent variable expression: η=Bη+Γξ+ζ η is the endogenous latent variable; B is the coefficient matrix between the endogenous latent variables, indicating the mutual influence between the endogenous latent variables; Γ is the coefficient matrix of the exogenous latent variable on the endogenous latent variable; ξ is the exogenous latent variable; ζ represents the error term of the structural equation.

5. The method for predicting travel route selection behavior based on survey and data analysis according to claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4.1, the model implicit covariance matrix is ​​as follows: Σ=Λy(IB) -1 GG T (IB) -T Yes T +Λy(IB) -1 Ψζ(IB) -T Yes T +Θ∈ Where Λy is the coefficient matrix of y to η, I is the identity matrix, B is the coefficient matrix between endogenous latent variables, Γ is the regression coefficient between exogenous latent variables and endogenous latent variables, Φ, Ψ, Θ∈ are the covariances of the errors between exogenous latent variables, between endogenous latent variable errors, and between observed variables, ζ represents the error term of the structural equation, and T represents the transpose of the matrix; Maximum Likelihood Estimation: FML=log∣Σ∣+tr(SΣ -1 )-log∣S∣-p In the formula, Σ represents a covariance matrix, |Σ| represents the determinant of Σ, log|Σ| is the logarithm of the determinant, S represents the sample covariance matrix, Σ -1 is the inverse matrix of Σ, tr(SΣ -1 ) is S and Σ -1 The trace of the product, p is the number of observed variables; Weighted least squares function: F WLS =(s-σ) T W(s-σ) Where s represents the vectorized elements of the sample covariance matrix; σ represents the vectorized elements of the model implicit covariance matrix; W represents the weighting matrix; Iteratively update the parameters according to the gradient of the objective function: In the formula, θ t is the parameter of the tth iteration, η is the learning rate, and F is the gradient; Use the indicator to evaluate the goodness of fit between S and Σ; If the fit is poor, modify the model and re-estimate; output estimation parameters include factor loading Λ, path coefficients (B, Γ), covariance (Φ, Ψ) and error variance Θ; fit index is an indicator of the degree of fit between the model and the data; residual matrix: the difference between S and Σ, highlighting the areas of model improvement; Step 4.2: Continue learning and feedback integration, use actual travel data for continuous training, and dynamically adjust predictions.

6. A method for predicting travel route selection behavior based on survey and data analysis according to claim 5, characterized in that: The modifying the model includes adding paths and adjusting constraints.

7. A system for predicting travel route selection behavior based on survey and data analysis, used to implement the method as claimed in claim 1, characterized in that: It includes data collection module, data analysis module, behavior analysis module, route selection optimization and recommendation module, and machine learning model module; The data collection module integrates data from multiple sources; The data analysis module prediction engine is driven by a machine learning algorithm and uses the collected data to predict average mileage and route selection behavior; The behavior analysis module analyzes factors that affect travel behavior; The route selection optimization and recommendation module provides optimized route recommendations to users based on predicted average mileage, estimated travel time, traffic congestion, and user preferences; The machine learning model module is trained on a large amount of travel history and traffic pattern datasets.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method of claim 1.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

Images