Iganfis Data Mining Approach for Forecasting Cancer Threats

Abstract

Healthcare facilities have at their disposal vast amounts of cancer patients’ data. Medical practitioners require more efficient techniques to extract relevant knowledge from this data for accurate decision-making. However the challenge is how to extract and act upon it in a timely manner. If well engineered, the huge data can aid in developing expert systems for decision support that can assist physicians in diagnosing and predicting some debilitating life threatening diseases such as cancer. Expert systems for decision support can reduce the cost, the waiting time, and liberate medical practitioners for more research, as well as reduce errors and mistakes that can be made by humans due to fatigue and tiredness. The process of utilizing health data effectively however, involves many challenges such as the problem of missing feature values, the curse of dimensionality due to a large number of attributes, and the course of actions to determine the features that can lead to more accurate diagnosis. Effective data mining tools can assist in early detection of diseases such as cancer. In This paper, we propose a new approach called IGANFIS. This approach optimally minimizes the number of features using the information gain (IG) algorithm which is usually used in text categorization to select the quality of text. The IG will be used for selecting the quality of cancer features by virtue of reducing them in number. The reduced number quality features dataset will then be applied to the Adaptive Neuro Fuzzy Inference System (ANFIS) to train and test the proposed approach. ANFIS method of training is ideally the hybrid learning algorithm which uses the gradient descent method and Least Square Estimate (LSE) for computing the error measure for each training pair. Each cycle of the ANFIS hybrid learning consists of a forward pass to present the input vector calculating the node outputs layer by layer repeating the process for all data and a backward pass using the steepest descent algorithm to update parameters, a process called back propagation.