Travel Prediction Using Spatial Time Feature Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current travel prediction methods rely on simplistic data mining techniques, primarily using user travel path and search data, resulting in inadequate prediction accuracy.
Innovation Solution
A travel prediction method that performs spatial and time information mining on search and travel path data, interacting the obtained features using a factorization machine model to predict user travel probabilities to target locations and time periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If model tree method or factorization machine method is used for travel prediction, then the prediction process is simple, but the prediction accuracy is insufficient
Solution Approach 1:
The patent segments the travel prediction process into distinct modules: spatial information mining module, time information mining module, feature interaction module, and prediction module. Each module processes specific aspects of the data independently, allowing for more sophisticated analysis while maintaining systematic organization and manageable complexity.
Solution Approach 2:
The patent introduces multiple dimensions of analysis by separately mining spatial information (location features) and time information (temporal features), then combining them through feature interaction. This multi-dimensional approach transforms the simplistic single-model prediction into a comprehensive multi-factor analysis system, significantly improving prediction accuracy.
2Device complexity
If only user search data and travel path data are used, then the data processing is simple, but the data mining is insufficient
Solution Approach 1:
The patent segments the data processing into two distinct mining processes: spatial information mining that extracts location-based features from search and travel path data, and time information mining that extracts temporal patterns. This segmentation enables deeper extraction of meaningful features from the same input data without proportionally increasing processing complexity.
Solution Approach 2:
The patent performs preliminary information mining to extract spatial and temporal features before the actual prediction process. By pre-processing the raw search and travel path data into structured spatial and temporal feature representations, the system prepares enriched input data that enhances subsequent prediction accuracy without adding complexity to the core prediction algorithm.
3Measurement precision
If spatial and time information mining is performed with feature interaction, then the prediction accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent divides the complex computational process into separate spatial mining, time mining, and feature interaction stages. Each stage processes specific types of information independently, allowing for optimized computation at each step rather than handling all features simultaneously, thus managing computational complexity more effectively.
Solution Approach 2:
The patent performs preliminary mining of spatial and temporal features before the prediction stage. By extracting and structuring these features in advance, the system reduces the computational burden during the actual prediction process, as the feature interaction module receives pre-processed, organized input rather than raw data requiring extensive processing.
Data Source
AI summary
A travel prediction method, apparatus, device, and storage medium. The method includes obtaining search data and travel path data of a user, performing spatial information mining based on the search data obtaining first spatial information and performing spatial information mining based on the travel path data obtaining second spatial information, performing time information mining based on the first spatial information obtaining first time information and performing time information mining based on the second spatial information obtaining second time information, performing information interaction based on the first spatial information, the second spatial information, the first time information, and the second time information obtaining an interaction feature, the interaction feature including a time feature and/or a spatial feature, and predicting a travel probability of the user going to a target location in a target time period based on the interaction feature, where the target time period matches the time feature, and the spatial feature matches the target location.