Multispectral Satellite Image Segmentation of Land Cover Area using Novel Naive Bayes in Comparison with Decision Tree

This work is based on the study of satellite images segmentation to find the land cover area accuracy based on Novel Naive Bayes which is compared with Decision Tree with improved accuracy in machine learning for classifying multispectral satellite image.For estimating the efficacy of the multispectral satellite image segmentation, the novel naive bayes algorithm with sample size of 20 and the decision tree algorithm with sample size of 20 were both calculated several times. Novel Naive Bayes have better accuracy compared to Decision Tree Algorithm. The power value is calculated using GPower 3.1 software (Gpower settings parameters with Gpower 80% for two groups; statistics analysis comparing two independent means; α=-0.05, power=-0.1) Novel Naive Bayes Algorithm has improved accuracy (92.5%) the standard deviation 0.05518 and the standard deviation error mean (0.01234) when compared with Decision Tree algorithm with improve accuracy (89.85%), The standard deviation (0.05518) and the standard deviation error mean (0.01323). The hypothesis is supported by the 2-tailed significance value of 0.002 (p<0.05), which is high. The results show that the Novel Naive Bayes Algorithm has segmented the image to find the land cover area and has got better accuracy and better performance compared to Decision Tree Algorithm by using the image segmentation process in machine learning.