1.3.1 Scalability:
Scalability is an issue of the algorithms with massive and real-world datasets. Because the analysis on core techniques progresses and matures, it becomes clear that a basic issue for RSs is to see the way to implant the core recommendation techniques in real operational systems and the way to contend with huge and dynamic sets of knowledge produced by the interactions of users with items. An answer that works fine once tested off-line on comparatively tiny knowledge sets may become inefficient might be or perhaps totally irrelevant on terribly massive datasets. New approaches and large-scale analysis studies are required [9].
1.3.2 Cold Start:
1.3.2.1 New User Problem:
It is a similar drawback with content-based and collaborative
…show more content…
Effective prediction of ratings from a little range of examples is very important. Also, the reliability of the collaborative recommendation system depends on the provision of a crucial mass of users. For example, within the movie recommendation system, perhaps there can be several movies that are rated by solely few individuals and these movies would be recommended terribly rarely, even if those few users gave high ratings to them. For the user whose tastes are uncommon compared to the remainder of the population, there will not be the other users who are particularly similar, resulting in poor recommendations.
1.4 Objective:
The main objectives of dissertation that are to be analyzed and can be implemented as follows:
1. The aim of recommendation system is to provide correct recommendations to user. The ‘Mean Absolute Error’ represents the effectiveness of results. The objective of work is to reduce MAE in comparison to traditional K-Nearest Neighbor algorithm.
2. Second objective is to give solution for ‘New user Problem’. To give solution for ‘new user problem’ user demographics will be analyzed.
3. Other objective of work is to give solution for ‘New Item Problem’. To give solution for ‘new item problem’ items demographics (genres) will be analyzed.
4. Also ‘sparsity’ of data should be