Svdadaptive Algorithm For Linear Least Square egressionandanomaly Reduction
Journal Title: IOSR Journals (IOSR Journal of Computer Engineering) - Year 2018, Vol 20, Issue 5
Abstract
In data analysis, the accuracy of analysisdependson (1) the data representation, (2) the algorithm and (3) the metric used to measure error. For ordinary linear least square approximation (OLA), the existing formulation measures error along y or vertical direction. Conventionally, ordinary linear least square approximation (OLA) technique has been considered as the best fit regression line for linear trend data. Based on domain knowledge, several versions of OLA have been developed. They are all reformulations of OLA using prior domain knowledge, for supervised learning. Singular Value Decomposition (SVD) is also used least square approximation. The robustness of SVD approximation is attributed to (1) the SVD line is sensitive to temporal variation in time variables whereas OLA is not, it makes OLA less suitable for time sensitive data, and (2) SVD has smaller approximation error than OLA regression line. But SVD has inherent weaknesses. Herein we present a hybrid algorithm that achieves a balance between quantitative and qualitative approximation accuracy of both OLA and SVD. This algorithm is also suitable for noise reduction. Visualization is a preferred way to ascertain the quality of a new algorithm, we use MATLAB R2017b and linear regression in simple two dimensional space to demonstrate the hybrid algorithm.
Authors and Affiliations
Chaman Lal Sabharwal
Keywords
Related Articles
- EP ID EP402134
- DOI 10.9790/0661-2005033348.
- Views 113
- Downloads 0
An Intelligent Approach to Information Retrieval System Using Enhanced DIG and FP-Tree Techniques
Abstract: Information retrieval is the process of retrieving all the relevant documents that satisfies the user query from large corpora. It is aimed to provide the relevant information and documents that matches the use...
A Novel Approach for Identification of Forest Fires using Land Surface Temperature Images
Abstract: Forest fire is those of the natural disasters that cause devastation to economy and environment. In order determine forest fire many of the detecting approaches have been applied. In this paper, we propos...
Smart Data Server for Smart Shops
The Now a day internet plays a very major role in business processes. The internet acts as the kind of the bonding between the customers and an organization (probably business), the customer can obtain the inform...
Challenges and Issues of Cluster Based Security in MANET
Abstract: Mobile ad-hoc networks are a specific kind of wireless networks that can be quickly deployed without pre- existing infrastructures. Even though there are various studies on cluster based security in MANET but t...
Support Vector Machine Based Sentiment Analysis Process for Twitter Streams
Abstract: Sentiment Research in twitter is quite tough due to its short length. Attendance of emoticons, slang words and misspellings in tweets compelled to have a preprocessing pace beforehand feature extraction. There...