Performance Analysis for Visual Data Mining Classification Techniques of Decision Tree, Ensemble and SOM

This research paper is a comprehensive report on experimental setup, data collection methods, implementation and result analyses of market segmentation and forecasting using neural network based artificial intelligence tools. The main focus of the paper is on visual data mining applications for enhancing business decisions. The software based system is implemented as a fully automated and intelligent enough to take into effect of each sales transaction. It updates and instantly modifies forecasting statistics by receiving input sales data directly from sales counter through networked connectivity. The connectivity may be wired or wireless. Three artificial intelligence tools, namely decision tree, ensemble classifier and Self Organizing Maps (SOM) are used for data processing and data analysis. The visual data mining concept is implemented by presenting results in the form of visual interpretation in as simple as possible way to understand very complex statistics. The current research results are mapped to interactive visualization by using multi-level pie charts, multi bar charts, histograms, scatter plots, tree maps and dataflow diagrams. The different visualization techniques help in understanding different levels of information hidden in very large data sets. The results analysis show that decision tree has classified data correctly up to a 86. 0 %, ensemble techniques produced an average of 88. 0 % and the predictions using SOM has accuracy of 90. 0 %. The survey carried out after implementation and use of the system shows that the system is very