Article Preview
Top1. Introduction
Intelligent systems are technically developed machines that deduce and induce the activation of the universe. Recommendation System is an intelligent system for learning resources. The basic model of recommendation algorithm is:
- 1.
The required features: This includes the users and the items.
- 2.
Two-dimensional Rating matrix: The mapping of users to the items.
- 3.
Selection of an item for recommendation: Recommend item (I) = Maximum rating of users to an item I.
Research has proposed various application areas of recommender systems, such as e-commerce, education, government, business, tourism, cloud service, shopping, medical, research, and collaboration field. Recommendation systems are also used in e-commercial web sites to provide the preferences to the user by analyzing user data and extracting patterns for further predictions. In the banking sector, customers may have complaints regarding the dealing procedure and working style followed by the employees of the bank. A large amount of data for the grievances can be analyzed to maintain effective CRM (Jain.et.al, 2016). Recommender System is facing many problems in today's dynamic environment such as issues related to scalability, sparsity, and Cold-Start Problem. Cold-Start Problem is the inability of the ratings contributed by the user to give a proper recommendation. The solution for the cold start problem is implicit ratings (Bauer.et.al, 2014). Sparsity and Scalability are managed without compensating for the accuracy of the recommendation process.
1.1 Working Principle of the Recommender System
We started with the study of the literature on Recommendation analysis procedures. Based on the various authors’ survey, we identified various approaches for analyzing information, characteristics of users and items for providing the recommendations. We collected database contents and applied the filtering algorithm to achieve the accuracy, precision and novelty of the predicted data. These recommendation models are then used for predicting the patterns to classify the categories. Figure 2 shows the working principle for a recommendation system.
Figure 2. Working road-map of recommender system
The Recommender system collects the ratings, user information, items features, and social user networks data and demographic information for filtering algorithm. Model-Based and Memory-Based selection techniques are probabilistic. To enhance the system performance in terms of memory and time consumption Nearest neighbors, Bayesian networks and Bioinspired algorithms are employed.
2. Classification Of Recommender System
A recommender system is classified based upon the attributes and characteristics required for the filtering process. Figure 3 shows the recommendation filtering process by taking the input of specific filtering techniques and provides the prediction of items to the user.
Figure 3. Filtering process for recommendation
According to (Burke.et.al, 2011), the models of various recommendation techniques are elaborated to understand the features of items and users.