Recommender Systems for Market Predictions

Abstract

The Internet's growth has resulted in a significant dispersion of data resources. Specialized recommendations on various types of information, products, and services are needed to support users in overcoming the problems of information overload. The recommendation system is one of the secrets ways utilized by successful companies, and this system considers a magical marketer for services and products b y observing customers and understanding their behavior to help them making the right decision. This thesis contains machine learning techniques combined with deep learning techniques. In addition, the apriori algorithm is used to create association rules because array based , large memory and it scans DB multiple times in order to improve the system efficiency to make accurate predictions and recommend suitable products. One of the most important techniques that was used in addition to apriori is a proposed model, this technique was applied by combining machine learning techniques and deep learning techniques by the application of a proposed model between GMM & KNN, GMM & SVM, GMM & LSTM. calculate (precision recall, f1-score, accuracy). The proposed approach utilized the Modcloth dataset sold Amazon containing of 998,94 transaction records. The obtained results were compared using assessment measures to determine which model is the best ,the results showed that the best classifier was K-means-SVM where achieved 0.999 rate of accuracy then the K-means-KNN achieved 0.998 rate of accuracy, based on the above results, the best classifier GMM-SVM where achieved the accuracy 0.996 , then the GMM-KNN achieved 0.992 rate . The proposed system was implemented using the Python programming language and imported some libraries to get high performance mode.