Clustering based Ensemble Machine Learning for Integrated Prediction and Classification of Ecological Zones

Accurate classification of ecological zones helps identifying the location of specific ecological features, and analyzing changing conditions. However, existing methods had limitations such as forming different image dimensions and analysing the areas within the images due to high noisy data. To overcome this, the proposed research integrated Density Based Spatial Clustering of Applications with Noise (DBSCAN) with Ensemble Machine Learning (EML) algorithms using Multilayer Perceptron (MLP), Naïve Bayes (NB), Classification and Regression Tree (CART), and Logistic Regression (LR). To train the model, the Landsat series data which is obtained from Geographical Information System (GIS) is used and then pre-processed using Normalized Difference Vegetation Index (NDVI) to characterize a variety of vegetation properties. The DBSCAN is then used for noisy data to distinguish the densities of data points and to segment data in different fields. The integration of DBSCAN with ML classifiers improved the classification accuracy. The overall experimental results show that the proposed DBSCAN-EML achieved 98.43% of accuracy and 97.89% of precision when compared with the existing methods EML based spatial prediction method and Artificial Neural Network Cellular Automata (ANN-CA).