Travel Way Recommendation Using Correlation-Based Vector Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for recommending travel ways struggle to accurately represent the heterogeneity among different travel modes, such as bus, car, bicycle, and walking, due to random initial vector generation, leading to suboptimal recommendations.
Innovation Solution
The method determines correlations between travel ways based on historical data, generates vectors representing these correlations, and uses learning algorithms to adjust initial user and starting-and-arrival pair vectors, ensuring that recommended travel ways reflect the similarities and differences among various modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If random initial vectors are generated for each travel way, then the system complexity is reduced and ease of manufacture is improved, but the manufacturing precision of vector representation deteriorates and cannot accurately represent heterogeneity among travel modes
Solution Approach 1:
The patent pre-calculates and stores correlation values between different travel ways before the recommendation process. These correlation values are computed based on historical travel data and stored in a correlation matrix, so that during actual recommendation, the system can directly use these pre-computed values to generate meaningful initial vectors without random generation, thereby accurately representing travel way heterogeneity from the start.
2Manufacturing precision
If correlation-based vectors are generated for each travel way, then the manufacturing precision of travel way representation is improved, but the device complexity and computation time increase
Solution Approach 1:
The system performs the computationally intensive correlation calculation in advance and stores the results in a correlation matrix. This preliminary action separates the heavy computation phase from the real-time recommendation phase, allowing complex precision vector generation without increasing the complexity of the operational recommendation system.
Solution Approach 2:
The patent creates a simplified correlation matrix that copies and stores the essential relationship patterns between travel ways. Instead of performing complex calculations during each recommendation, the system uses this pre-computed correlation matrix as a compact representation that can be efficiently queried and used to generate accurate initial vectors.
3Reliability
If learning is performed on initial vectors using historical travel data, then the reliability of travel way recommendations is improved, but the loss of time for data processing increases
Solution Approach 1:
The patent pre-processes historical travel data to extract correlation patterns between travel ways and stores them in a correlation matrix before the actual recommendation need arises. This preliminary data processing separates the time-consuming learning phase from the time-critical recommendation phase, allowing the system to achieve high reliability through thorough learning while maintaining fast response times during actual use.
Data Source
AI summary
The present disclosure provides a method and an apparatus for recommending a travel way. The method includes: obtaining historical travel data; determining a plurality of correlations between respective travel ways according to the historical travel data; generating a vector corresponding to each travel way according to the plurality of correlations between respective travel ways; performing learning on an initial vector corresponding to each user and an initial vector corresponding to each starting-and-arrival pair according to the vector corresponding to each travel way and the historical travel data, to obtain a vector corresponding to each user and a vector corresponding to each starting-and-arrival pair; and recommending a travel way according to the vector corresponding to each travel way, the vector corresponding to each user and the vector corresponding to each starting-and-arrival pair.