Graph based approach for minimum multicollinearity highly accurate regression model explaining maximum variability

Regression is a supervised machine learning method which advances by building a equation by which we can estimate the response variable. The proper choice and number of predictors used for prediction of response variable plays an important role. Multicollinearity affects the model accuracy hence emp...

Full description

Saved in:
Bibliographic Details
Main Authors: Sharma, Mrityunjay, Saha, Suman
Format: Conference Proceeding
Language:English
Subjects:
Accuracy
Buildings
Correlation
Equations
Information technology
Mathematical model
Mean Square error
Multicollinearity
Predictive models
PRESS residual
Regression
Variance Inflation Factor
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Regression is a supervised machine learning method which advances by building a equation by which we can estimate the response variable. The proper choice and number of predictors used for prediction of response variable plays an important role. Multicollinearity affects the model accuracy hence emphasis has to be made to keep it as low as possible. The coefficients of the regression equation are calculated by X T X. In case of strong correlation amongst attributes it is difficult for software to calculate this term. In addition to all this the approach has to be pictorial or graphical in nature for better comprehensibility which is unique. This paper keeps in mind all these factors and proposes a technique for building a best (accurate, minimum multicollinearity, graphical) regression model.
DOI:10.1109/CONFLUENCE.2014.6949292