Adaptive Recommendation System Using Time-Weighted Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for recommending future purchases to consumers lack the ability to incorporate a time aspect, leading to irrelevant recommendations over time and requiring access to personal information.
Innovation Solution
An adaptive recommendation system that uses a computer device to receive preference vectors based on payment transactions, incorporates a time aspect into consumer actions, and generates personalized vectors to provide tailored recommendations without transmitting personal information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If known systems use central servers to access customer personal information for recommendations, then recommendation coverage can be achieved, but recommendation accuracy deteriorates over time due to lack of time aspect incorporation
Solution Approach 1:
The system dynamically adjusts recommendation relevance by incorporating time aspects into consumer action analysis. The AR computer device processes time-stamped consumer actions and updates preference vectors dynamically, allowing recommendations to adapt to changing consumer behavior patterns over time rather than using static historical data.
Solution Approach 2:
The system changes the parameter of time incorporation in recommendation algorithms. By adding time aspect parameters to consumer action data and using these to weight and filter preference vectors, the system transforms static recommendation models into time-aware models that maintain accuracy by considering when actions occurred.
2Productivity
If known systems continuously transmit personal information to central servers, then centralized processing can be achieved, but communication traffic increases and personal information security deteriorates
Solution Approach 1:
The system extracts and processes only necessary aggregated preference data locally at the AR computer device rather than continuously transmitting all personal information to central servers. The AR device receives consumer actions, determines personalized preference vectors locally, and only transmits minimal recommendation results, eliminating unnecessary communication traffic while maintaining processing efficiency.
Solution Approach 2:
The AR computer device performs self-service by locally processing consumer action data and generating recommendations without requiring continuous central server intervention. The device autonomously determines personalized vectors and filters recommendations based on time aspects and current context, reducing dependency on centralized processing and minimizing communication overhead.
3Device complexity
If known systems make recommendations based on past purchases without time aspects, then simple processing can be achieved, but recommendation relevance deteriorates as connections become outdated
Solution Approach 1:
The system performs preliminary action by pre-processing consumer actions into time-stamped data structures and pre-calculating preference vectors with time weights before they are needed for recommendations. This preparation work is done in advance so that when recommendations are needed, the system can quickly retrieve and filter pre-processed data rather than performing complex time-based calculations in real-time.
Data Source
AI summary
A computer-implemented method for recommending items for future purchase by a consumer based on the consumer's historical actions is provided. The method is implemented using an adaptive recommendation (“AR”) computer device in communication with a memory. The method includes receiving from a recommender server computing device at least one preference vector, storing the at least one preference vector, receiving at least one consumer action including a time aspect, from a candidate consumer, and determining at least one personalized vector based on the at least one preference vector, the at least one consumer action, and the time aspect associated with each of the at least one consumer action. The at least one personalized vector represents at least one future purchase that the candidate consumer will likely conduct. The method also includes displaying at least one recommendation to the candidate consumer based on the at least one personalized vector.