Recommendation System Using Preference Correlation and Reservation Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional recommendation systems using content-based filtering are limited to items understandable by computers and rely on pre-set keywords or rules, while cooperation filtering requires extensive user evaluation, leading to a burden on users and limited recommendations to previously evaluated items, including unbroadcasted content.
Innovation Solution
An information processing system calculates a predicted preference value for new items based on reservation information, allowing for the recommendation of unevaluated items by transmitting user preference values and reservation data through a network, enabling the presentation of recommended items to users without prior evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If content-based filtering is used to select items based on keywords and rules, then the selection process is automated and efficient, but the selected items are limited to those that can be understood by a computer and match pre-set keywords
Solution Approach 1:
The patent introduces a mediator component that translates user preferences and item characteristics into a common representation space, enabling the system to handle both structured (keyword-based) and unstructured (content-based) information. This intermediary layer allows automated processing while expanding the range of selectable items beyond predefined keywords.
Solution Approach 2:
The system dynamically adjusts selection parameters by transforming rigid keyword matching into flexible similarity scoring. By changing the parameter from binary keyword presence/absence to continuous similarity scores, the system maintains automation while significantly expanding the versatility of selectable items.
2Adaptability or versatility
If cooperation filtering is used to recommend items based on user group preferences, then items can be selected from unspecific items, but many users must evaluate plural items in advance which imposes a heavy load on users
Solution Approach 1:
The system performs preliminary actions by pre-calculating user preference profiles and item characteristics during off-peak times. This allows the recommendation engine to quickly match users with suitable items without requiring users to perform real-time evaluations, thus reducing the operational burden while maintaining versatile recommendations.
Solution Approach 2:
The patent implements self-service mechanisms where the system automatically gathers and processes preference data from user interactions without requiring explicit user evaluations. The system serves itself by learning from implicit feedback (viewing history, purchase behavior) to generate recommendations, eliminating the need for users to manually evaluate items.
3Device complexity
If items are selected from given selection items based on predetermined rules, then the recommendation process is straightforward, but the recommended item can be selected only from among the given items
Solution Approach 1:
The patent adds another dimension to the recommendation process by introducing a dynamic item generation component. Instead of selecting only from existing items, the system can create new item representations by combining features from multiple existing items, thus maintaining process simplicity while expanding selection flexibility to include synthesized items.
4Reliability
If conventional cooperation filtering is used, then recommended items are limited to items already evaluated by other users, but this prevents recommendation of unevaluated new items such as unbroadcasted programs
Solution Approach 1:
The system performs preliminary analysis of new items by comparing their characteristics with previously evaluated items. By pre-processing new items through feature extraction and similarity computation against the existing item database, the system can generate recommendations for unevaluated items while maintaining reliability through comparison with established evaluation patterns.
Solution Approach 2:
The patent introduces an intermediary mapping mechanism that connects new, unevaluated items to the existing evaluation framework. This mediator translates new item characteristics into the same feature space as evaluated items, allowing the system to leverage existing user preference data to recommend new items without requiring prior user evaluations of those specific items.
Data Source
AI summary
An unbroadcast program is recommended to a user. A server is supplied with degree of preference data as indices of the taste of users for programs from user terminals, and calculates, based on this, correlations of the preference tendencies for the programs of other users using a service. The user terminal displays an electronic program table transmitted from the server, and when a reservation to view or record is input from the user, reservation information is transmitted to the server. The server calculates predicted values of degrees of preference of the respective users for programs to be broadcast in the future on the basis of the correlations of the degrees of preference and the reservation information, and transmits them to the respective user terminals. The user terminal refers to the degree of preference predicted values of the broadcast scheduled programs, and prepares a list of programs to be recommended to the user.