Hybrid Recommendation System for Islamic Dari-Textual Information Context using Machine Learning and AI

Life in the world of massive data is too challenging. Although, more information is good but few relevant information brings better decision, quality, and simplicity in life. Islam's religion has considerable amount of important textual information in the format of books and papers in all areas of life. Therefore, through modern technology, personalizing and prioritizing information retrieval or filtering techniques should be used. And by utilizing of these techniques, we can effectively overcome the information overload problem. Implementation of all aforementioned techniques in a single system and develop a compact pipeline of data is called recommendation system (RS). In this case we developed five individual RS models such as (Content-based, Sequence-based, Deep Neural Network collaborative filtering: DNN CF Model, DNN CF plus descriptive attributes Model and finally popularity-based) with the capability of 10 ways to individually generate recommendation, and 45 possible ways to combine two RS together to improve power of their final recommendation which is called hybridization or assembling Methods. Finally, evaluation of our individual trained models and weighted hybridized forms are done on our mini dataset, as well as on 100K movielens global dataset. 

In this research, we have sufficient overview of Islamic Information context and fields of knowledge.