Intelligent Recommender System Based on Machine Learning Techniques

نوع: Type: thesis

مقطع: Segment: masters

عنوان: Title: Intelligent Recommender System Based on Machine Learning Techniques

ارائه دهنده: Provider: Mohammad Jamali Dogaheh

اساتید راهنما: Supervisors: Dr. Hassan Khotanlou

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees:

زمان و تاریخ ارائه: Time and date of presentation: 2020

مکان ارائه: Place of presentation:

چکیده: Abstract: Nowadays internet users have access to a lot of information and choices. It is not an easy task for users to search all of the available data to find their needs. Recommender systems (RS) invented to explore all products and recommend the products that users need. Recommender systems reconstruct the user preference matrix using dimensionality reduction. The overall process of reconstruction consists of transforming user preferences to latent space and reversing the transformation. However, using traditional approaches end up with acceptable prediction results but they are still suffering from problems like cold start and sparsity. "Factorization metric learning" uses Euclidean distance as a transform and inverse transform function. Euclidean distance is noise sensitive and it's not accurate when variables are correlated. Mahalanobis distance (MD) uses the covariance matrix to reduce the correlation between variables and measure the distance correctly. We noticed that factor analysis approaches are highly parameter sensitive so using Mahalanobis distance directly would not help to improve the prediction. In this research, we study using Euclidean distance as transform function and a simplified Mahalanobis distance which is equal to a weighted Euclidean distance alongside feature crosses as inverse transform function. Also, a customized neural network has been studied which exploits the Euclidean distance feature in its first layer as an invert of transform function. Key Words: Recommender System, Machine Learning, Latent Space, Euclidean Distance, Mahalanobis Distance, Neural Network

فایل: ّFile: Download فایل