期刊名称：Indian Journal of Computer Science and Engineering
印刷版ISSN：2231-3850
电子版ISSN：0976-5166
出版年度：2021
卷号：12
期号：2
页码：348-361
DOI：10.21817/indjcse/2021/v12i2/211202047
出版社：Engg Journals Publications
摘要：Rainfall detection becomes an important task during varying climatic conditions. It is become very essential to examine the changing patterns of the rainfall and try to detect the rain. The rainfall databases are highly susceptible to noise, inconsistent and missing of data at present day in the realworld. So, the data obtained must be repaired by plugging absent values and eliminating the unrelated data. For this min-max normalization is employed for transferred the data into [0, 1] range. After that it was operated with z-score normalization, where the data values are distributed in a small scope. Then the feature extraction using PCA and ACO is functioned. Finally, naive bayes classification is applied to compare the performance of with and without pre-processing and the classes representation based on the different principal components (PC) of feature selection were calculated with and without pre-processing. Monthly rainfall data sets that are downloaded from Indian meteorological department (IMD). Monthly rainfall for years 1901 to 2019 are taken for analysis. The missing data and the outliers were detected every month. For investigation, the value distributions and features using PCA and ACO of rainfall feature before and after normalization is observed. The min-max normalization and z-score normalization concluded that they do not change the autocorrelation. The ACO based method found efficient when compared to PCA method. Classification error with and without data pre-processing is measured and found that the pre-processing rainfall data enhances the model performance.
