Conceptual Model for Predictive Analysis on Large Data

Conceptual Model for Predictive Analysis on Large Data

DOI: 10.4018/978-1-5225-5029-7.ch003

Abstract

This chapter provides an overview of the proposed model for pattern extraction and pattern prediction over data warehouses. As discussed before, the main objective of the research is to provide a single model for pattern extraction and prediction. The objectives include an automated way to select variables for the mining process, automated schema design, advanced evaluation of extracted patterns, and visualization of extracted patterns.
Chapter Preview
Top

3.1 Proposed Methodology: Framework And Dataset Introduction

Figure 1 shows the model showing different phases involved in the model. The extraction and prediction of patterns is shown by using different paths. The extraction of patterns is shown through path E in the figure. The prediction of patterns is done through Path P1, P2 and P3 depending upon the nature of variables being used for prediction and the nature of variables being predicted.

Path E in the Figure 1 contains steps to extract patterns. The steps involve generation of clusters, ranking of variables, multidimensional scaling, STAR schema generation, mining of association rules, evaluation using advanced measures and visualization of association rules. Pattern prediction is done using paths P1, P2, and P3. Path P1 in the figure contains steps to predict nominal variables. This path involves generation of clusters, ranking of variables, building of classification model, evaluation of models, selection of model and prediction using the selected model. Path P2 in the figure shows the steps to predict numeric variables. The involved steps are generation of clusters, ranking of variables, STAR schema generation, classification model building, evaluation of models, selection of model and prediction of numeric variables. Path P3 in the figure shows the steps to predict aggregate data in the data warehouse environment. These steps involve generation of clusters, ranking of variables, STAR schema generation, data cube construction, building of classification models, evaluation of models, selection of a model and prediction of aggregate data.

Figure 1.

Model for multi-level pattern extraction and pattern prediction

Consider a real world mixed variables data set D of Movies, Concerts and TV Shows from IMDB database having four numerical and five nominal variables with 432 records in it. Numeric variables include User Rating, No. of Votes, US Box Office Gross, and Cast (number). The nominal variables include Title Type, Genres, Title Groups, Company, Production Status. Nominal variables are given in the Table 1 along with their respective distinct values.

Table 1.
Distinct Values for nominal variables in the example dataset D
VariablesDistinct Values
Title TypeFeature Film, TV Movie, TV Series, TV Episode, TV Special, Mini-Series, Documentary, Video Game, Short Film
GenresAction, Adventure, Animation, Biography, Comedy, Crime, Documentary, Drama, Family, Fantasy, Film-Noir, Game-Show, History, Horror, Music, Musical, Mystery, News, Reality-TV, Romance, Sci-Fi, Sport, Talk-Show, Thriller, War, Western
Title GroupsNow-Playing, Oscar-Winning, Best Picture-Winning, Best Director-Winning, Oscar-Nominated, Emmy Award-Winning, Emmy Award-Nominated, Golden Globe-WinninG, Golden Globe-Nominated, Razzie-Winning, Razzie-Nominated, National Film Board Preserved
Company20th Century Fox, Sony, DreamWorks, MGM, Paramount, Universal, Walt Disney, Warner Bros
Production StatusReleased, Post-production, Filming, Pre-production, Completed, Script, Optioned Property, Announced, Treatment/outline, Pitch, Turnaround, Abandoned, Delayed, Indefinitely Delayed, Active, Unknown

Complete Chapter List

Search this Book:
Reset