DESIGN AND DEVELOPMENT OF MISSING VALUES AND PREDICTION OF TIME SERIES DATA

Abstract

Data preprocessing plays a huge and essential capacity in the data mining measure. Data preprocessing is expected to improve the adequacy of a figuring. This paper bases on missing worth evaluation and estimate of time course of action data subject to the undeniable characteristics. Different figurings have been made to handle this issue, yet they have a couple of limitations. Most existing counts like KNNimpute (K-Nearest Neighbors attribution), BPCA (Bayesian Principal Component Analysis) and SVDimpute (Singular Value Decomposition credit) can't deal with the condition where a particular time point (portion) of the data is missing inside and out. This paper revolves around autoregressive-model-based missing worth appraisal technique (ARLSimpute) which is ground-breaking for the situation where a particular time point contains many missing characteristics or where the entire time point is missing. Data preprocessing yield is given to the commitment of the desire systems to be explicit direct conjecture and quadratic figure. These techniques are used to envision the future characteristics reliant on the chronicled values. The introduction of the figuring is assessed by execution estimations like precision and audit. Test results on certified datasets show that the proposed figuring is feasible and viable to reveal future time course of action data