WiMi Developed a Machine Learning-based Multi-modal Fusion Recommendation System

WiMi Hologram Cloud a leading global Hologram Augmented Reality (“AR”) Technology provider announced that a multi-modal fusion recommendation system based on machine learning is developed to provide users with more accurate and diverse recommendation results by fusing multi-modal data from multiple interaction types and attribute modes, which is currently mainly applied in the field of e-commerce.

AiThority Interview Insights : AiThority Interview with Babak Pahlavan, Founder and CEO of NinjaTech AI

WiMi evaluates the performance of a multi-modal fusion recommendation system by conducting a series of experiments on an open dataset. The experimental results show that WiMi’s machine learning-based multi-modal fusion recommendation system achieves significant improvements over the benchmark results of the latest existing work. Practical use cases of our system are also demonstrated on several e-commerce websites. Taking food and beverages, shoes, fashion items and telecommunication carriers as examples, our system was successfully applied in these domains and provided users with an accurate and personalized recommendation experience. By analyzing user behavioral data such as clicks, purchases, add-to-cart, etc., and combining information from multiple attribute patterns, our system is able to accurately recommend relevant products, helping users find what they need and make faster purchasing decisions.

Technical logic of a machine learning-based multi-modal fusion recommendation system of WiMi:

Data representation and pre-processing: The first step in a multi-modal fusion recommendation system is to collect and pre-process data. User behavior data such as click, buy, add to cart, etc. are obtained from various interaction data sources. At the same time, data from multiple attribute modalities such as audio, video, image and text are collected. These data are pre-processed, feature extracted and cleaned to prepare for subsequent data fusion and model training.

Multi-modal data fusion: Multi-modal data fusion is the core technology of the system. It utilizes deep learning models and graph embedding algorithms to transform data from different attribute modalities into unified vector representations. By fusing these vectors, the correlation and similarity between different attribute modes can be captured, thus realizing cross-modal data fusion.

Intelligent recommendation algorithm: A deep learning network for intelligent recommendation is trained based on fused multi-modal data representations. The network utilizes data from multiple interaction types for model training and optimization to generate personalized recommendation results. The system uses visual data embedding and efficient graph embedding algorithms to enhance the performance and effectiveness of the recommendation algorithms. These algorithms can effectively mine and utilize the rich information of multi-modal data to provide more accurate and diverse recommendation results.

Read More about Interview AiThority: AiThority Interview with Gijs van de Nieuwegiessen, VP of Automation at Khoros

Business rules and real-time adjustment: WiMi’s multi-modal fusion recommendation system allows users to define and adjust business rules to fit different recommendation scenarios and needs. By parsing and running the business rules, accurate recommendation results can be generated according to specific business logic. At the same time, the system is also equipped with real-time adjustment capability, which allows the recommendation algorithm to be dynamically adjusted and optimized based on experimental and measurement results. This ensures that the recommendation system is always efficient and accurate.

WiMi’s machine learning-based multi-modal fusion recommendation system provides an efficient and intelligent recommendation framework by utilizing data fusion of multiple interaction types and attribute modes. Through the combination of deep learning models, graph embedding algorithms and business rules, enough to generate accurate, personalized recommendation results with real-time adjustment and optimization capabilities, the system can be successfully applied to different e-commerce fields. Besides, WiMi’s multi-modal fusion recommendation system can also be applied in social media, video streaming, travel and hotel, online education and other fields.

For example, in the field of social media, social media platforms can use the multi-modal fusion recommendation system to recommend more interesting and personalized content for users, enhancing user retention and engagement. The system can combine data from users’ social behavior, text content, images and videos to provide users with content recommendations related to their interests and preferences, improving user experience and platform activity. In the field of video streaming, the multi-modal fusion recommendation system can provide more intelligent and personalized video recommendations on the video streaming platform. By combining the user’s viewing behavior, video content, audio and other data, the system can recommend the video content they may be interested in, improving the user’s viewing experience and the platform’s user retention rate. In the field of online education, the multi-modal fusion recommendation system can provide smarter and personalized learning resource recommendations on the online education platform. By combining students’ learning behaviors, text content, and audio and video data, the system can recommend teaching resources that meet students’ learning needs and interests, and improve students’ learning effectiveness and satisfaction.

In the future, WiMi will continue to improve and optimize our multi-modal fusion recommendation system. We plan to further improve the efficiency and accuracy of our data processing and fusion algorithms, and explore more advanced deep learning models and embedding algorithms to provide more accurate and diverse recommendation results. At the same time, WiMi will also strengthen its support for business rules and its ability to adjust dynamically to meet changing recommendation needs and scenarios. It is believed that through continuous innovation and technological advancement, WiMi’s multi-modal fusion recommendation system will provide users with a better and more personalized recommendation experience and bring greater business value to e-commerce and other fields.

Latest AiThority Interview Insights : AiThority Interview with Keri Olson, VP at IBM IT Automation

 [To share your insights with us, please write to sghosh@martechseries.com]