Collaborative filtering: special case in predictive analysis

Abstract

The data-flow across the internet has come to a point where the human brain is unable to read, analyze, and conclude the hidden knowledge. One way of helping people in their decisions is the classical query search and filters plugged into various available software in use. Still, the last decision is left to humans where the search for the best alternative is not provided by these queries.The need for a sort of an electronic learner has emerged. Such a type of learner has the main task of deciding for the user which preference (decision) he/she is likely to possess on the basis of other users' preferences (decisions). This field of analysis is called collaborative filtering. Learners have varied in performance and thus in acquiring results on the basis of the different algorithms involved.This article introduces the collaborative filter methods, also called recommender systems. Few publications and research have been elaborated on this topic, but the available studies are encouraging.

Network-Based Marketing: Identifying Likely Adopters via Consumer Networks

“…In general, the empirical studies that test and extend accepted theories of product diffusion rely on aggregate-level data for both the customer attributes and the overall adoption of the product (Ueda, 1990;Tout, Evans and Yakan, 2005); they typically do not incorporate individual adoption. Models of product diffusion assume that network-based marketing is effective.…”

Feature reduction for product recommendation in internet shopping malls

“…There have been a large number of studies for recommendation of products to promote cross-selling or up-selling in Internet shopping malls. The studies can be broadly classified into two types: collaborative filtering methods that use similarity of ratings for products among shoppers, and content-based filtering methods that utilize features or attributes of products and users [2,3,7,10,15,16]. Although collaborative filtering methods have been proved to be successful in many studies, it is often very difficult or expensive to collect the ratings, and moreover, when rating data is sparse, recommendation results are usually impaired severely [9,14,18].…”