Classifying crop pest data using C4.5 algorithm

摘要

Data mining is a way of exploring large preexisting databases in order to generate new information. It is used to find a relationship between the bulky data set which is very helpful in decision making. In agriculture sector, data mining plays an emerging role. Various data mining techniques can be used to protect crops from vertebrate pests, diseases so as to enhance risk on crop cultivation. This paper comprises data pre-processing to remove noisy data in crop pest data that offers better accuracy. Feature selection takes an essential pre-processing step is to reduce the cost of learning by reducing the number of attributes. In this paper Relief and Random Forest Filters are applied for filtering crop pest data set attributes instead of using full attribute set. Relief carries out a selection of instances randomly for calculating the attribute weights. Random forest retains random selection, but provides two straightforward methods such as mean decrease impurity and mean decrease accuracy. Depending upon weights, splitting attributes have been chosen for generating decision tree. This paper proposed C4.5 algorithm that handles crop pest training data with missing values and eliminates overfitting while construction of the tree, that improves the accuracy of the algorithm.