Source-linked AI summary
Normalization: A Preprocessing Stage
S. Gopal Krishna Patro, Kishore Kumar Sahu
TL;DR
Normalization is used to rescale data before subsequent processing, particularly in fields such as soft computing and cloud computing. The paper proposes Integer Scaling Normalization based on existing methods and compares it with Min-Max normalization across several datasets. The proposed technique produces normalized values between 0 and 1 and is presented for use with structured datasets and further research applications.
Problem
Data used in research fields may be unstructured, motivating normalization to scale or transform datasets before further processing.
Method
The paper proposes Integer Scaling Normalization, described as an AMZD-based technique that structures or transforms datasets into a normalized form.
Results
The proposed normalization technique produces values between 0 and 1 and is compared with Min-Max normalization using BSE Sensex, NNGC, and college enrollment datasets.
Takeaways & Limitations
The paper presents the technique as applicable to research areas including soft computing, image processing, cloud computing, and time-series financial forecasting.
Abstract
from arXiv · showhide
As we know that the normalization is a pre-processing stage of any type problem statement. Especially normalization takes important role in the field of soft computing, cloud computing etc. for manipulation of data like scale down or scale up the range of data before it becomes used for further stage. There are so many normalization techniques are there namely Min-Max normalization, Z-score normalization and Decimal scaling normalization. So by referring these normalization techniques we are going to propose one new normalization technique namely, Integer Scaling Normalization. And we are going to show our proposed normalization technique using various data sets.
I. INTRODUCTION
Normalization is presented as a preprocessing and mapping technique that rescales data into a new range for prediction and forecasting. The paper introduces Integer Scaling as a new technique alongside established methods.
- Normalization rescales or maps an existing data range into a new range before further processing.
- The technique is described as useful for prediction and forecasting tasks with substantial variation.
- The paper identifies Min-Max, Z-score, and Decimal scaling as existing normalization techniques.
- Integer Scaling is proposed as a new normalization technique derived from AMZD.
II. RELATED STUDY
The related study reviews Min-Max, Z-score, and Decimal Scaling normalization methods and prepares the comparison with the proposed technique. These methods transform unstructured data into normalized ranges using different operations and parameters.
- The paper states that the proposed technique will be discussed in detail after reviewing these existing methodologies.
- Min-Max normalization linearly transforms original data into a predefined boundary while preserving relationships among values.
- Z-score normalization uses the mean and standard deviation to calculate normalized values for each row and column.
- Rows with identical values have zero standard deviation, so their Z-score-normalized values are set to zero.
- Decimal Scaling is described as producing values in the range -1 to 1, with the scaling exponent chosen so Max(|v_i|)<1.
III. PROPOSED MODEL
The proposed AMZD normalization transforms integer data into the range 0 to 1 and is intended to structure datasets for subsequent processing. It operates on individual elements, regardless of dataset or element size, and is compared with Min-Max normalization across three datasets.
- The proposed technique transforms unstructured or poorly structured datasets into normalized data within the range 0 to 1.
- AMZD normalization provides values between 0 and 1, like Min-Max, Z-score, and Decimal Scaling normalization.
- The technique scales or transforms each individual element independently of dataset size and the number of digits in each element.
- AMZD normalization is applicable only to integer numbers and can process integer elements of any length.
- The comparison study evaluates the proposed technique against Min-Max normalization using BSE Sensex, NNGC, and college enrollment datasets with tables and graphs.
- The normalization workflow selects a data range, reads it, scales it to between 0 and 1, applies the result in further processing, and scales it up if required.
IV. CONCLUSION
The paper states that the proposed normalization technique works across fields including soft computing, image processing, and cloud computing. It also proposes future use in time-series financial forecasting and other normalization techniques.
- The authors state that the normalization technique works in soft computing, image processing, and cloud computing.
- The authors plan to develop other normalization techniques and apply this technique to time-series financial forecasting.
BIOGRAPHY
The authors’ biographies describe their academic roles, research experience, and areas of interest in computing.
- S.Gopal Krishna Patro is a M.Tech. research scholar in CSE & IT at VSSUT, Burla, specializing in ICT.
- His research interests include financial forecasting, machine learning, and cloud computing.
- Patro has published one international journal article and has two years of combined industry and teaching experience.
- Kishore Kumar Sahu is an Assistant Professor in the CSE & IT department at VSSUT, Burla, with ten years of undergraduate and postgraduate teaching experience.
- Sahu is pursuing a Ph.D. in Computer Science and Engineering and researches soft computing, artificial intelligence, compiler, and theory of computation.