Informatics Publishing Limited

G. Purusothaman ¹, P. Krishnakumari ¹

Affiliations
1 Department of Computer Applications (MCA), RVSCAS, Coimbatore, 641402, IN

Abstract

Comparison of classification techniques in Data mining to find the best technique for creating risk prediction model of heart disease at minimum effort. In Data mining, different methods used to find risk prediction of heart disease. There are two types of model used in analysis of data. First one is applying single model to various heart data and another one is applying combined model to the data. The combined model also known as hybrid model. This paper provides a quick and easy understanding of various prediction models in data mining and helps to find best model for further work. This is unique approach because various techniques listed and expressed in bar chart to understand accuracy level of each. These techniques are chosen based on their efficiency in the literature. In previous studies of different researcher expressed their effort on finding best approach for risk prediction model and here we found best model by comparing those researcher’s findings as survey. This survey helps to understand the recent techniques involved in risk prediction of heart disease at classification in data mining. Survey of relevant data mining techniques which are involved in risk prediction of heart disease provides best prediction model as hybrid approach comparing with single model approach.

Keywords

Classification, Data Mining Algorithms, Heart Disease Risk Prediction, Hybrid Techniques, Survey of Data Mining Techniques, Prediction Models

Full Text

   

G. Purusothaman ¹, A. Nithya ²

Affiliations
1 Rathnavel Subramaniam College of Arts and Science, Sulur, Coimbatore - 402, Tamil Nadu, IN
2 Department of CS, Rathnavel Subramaniam College of Arts and Science, Sulur, Coimbatore - 402, Tamil Nadu, IN

Abstract

Heart disease is the frequently found disease in various peoples which would cause more serious and dangerous effects. Various studies have been projected earlier whose major aim is to predict the heart disease more accurately. In our previous research method Fuzzy Rough Set Theory combined with Support Vector Machine (FRS - SVM) is introduced which can ensures the optimal prediction rate by selecting the risk factors accurately which can lead to improved accuracy rate. However FRS-SVM might lack in its performance in case of presence of more missing values in the database. This research method cannot support the large dimensional dataset which needs to be focused well enough for accurate prediction rate. This problem is resolved in this investigation by introducing the framework namely Heart disease prediction using Alpha Rough Set Theory combined with Fuzzy SVM (ARST-FSVM). In this research method, Modified K-Means clustering algorithm is utilized for preprocessing the input dataset which would avoid the noisy data present in the database. Then missing data value in the database is handled using normalization technique where NLLS imputation is applied. And then feature dimensionality reduction is done using Alpha rough set theory (α-RST) approach. From those reduced feature set, optimal feature selection in terms of relevancy is done using Hybrid Bee colony algorithm with Glowworm Swarm Optimization (HBC-GSO) approach. Finally heart disease prediction is done using classifier namely fuzzy based SVM. The overall research method ensures that the proposed research technique leads to ensure it can direct to most favorable and accurate heart disease diagnosis outcome.

Keywords

Large Data Set, Heart Disease Prediction, Missing Values, Accurate Observation, Feature Reduction.

Full Text

     

V. Yuvaraj ¹, G. Pandiyan ¹, G. Purusothaman ¹

Affiliations
1 School of Computer Studies, Rathnavel Subramaniam College of Arts and Science, IN

Abstract

Lung cancer is one of the fatal forms of cancer worldwide. Genetic variability has been identified as influencing a person vulnerability to lung cancer in epidemiologic research. A new study undertaken by a team of experts from the United States National Cancer Institute on 14,000 Asian women discovered that Asian women, regardless of whether they smoke or not, are more likely to acquire cancer owing to genetic abnormalities. Early detection of this lethal disease is a novel clinical application of microarray data. Recent research establishes a model for the early diagnosis of lung cancer. Additionally, multilayer perceptron, random subspace, and Sequential Minimal Optimization (SMO) approaches are used for classification. While information acquisition is typically a good indicator of an attribute significance, it is not perfect. A noticeable issue develops when knowledge gain is applied to qualities that might take on many distinct values. This paper provides an efficient gene selection model based on the Improved Whale Optimization Algorithm (IWOA) to address these concerns. It saves time and identifies relevant genes from gene expression data, increasing lung cancer categorization accuracy. Then, a Modified Long Short-Term Memory (MLSTM) Network is used to classify lung cancer. It accepts specified genes as inputs and determines which class they belong to, such as lung cancer or normal subjects. As demonstrated by empirical observations, the suggested model is effective in precision, recall, accuracy, and f–measure.

Keywords

Lung Cancer, Early Stage, Developing Cancer, Genetic Variations, Feature Selection, Information Gain Attribute, Whale Optimization, Long Short Term Memory

Full Text