Data Mining Methodology

Chinky Gera

Department Of Computer Science & Engineering

RIMT Institute of Engineering and Technology

Mandi Gobindgarh, India [e-mail] Kirti Joshi

Department Of Computer Science & Engineering

RIMT Institute of Engineering and Technology

Mandi Gobindgarh, India [e-mail] Abstract— Patient’s post operative data is allied with hypothermia which is a serious concern after surgery. To operate against hypothermia is crucial

It is an inter-disciplinary field which integrates the data, machine learning, database technology and artificial intelligence. Data mining is an application of automatically searching the huge reserve of data to uncover the patterns and trends that go beyond easy analysis. Data mining plays an important role in the healthcare industry which enables the health systems to recognize the inefficiencies as well as practices to improve care by systematically using the data and analysis. The processes have been developed by the health systems so that the patient’s receive suitable care or treatment.[A] Medical databases are ever-growing information in data mining field collected from hospitals about patients and their medical conditions. The dataset occur from the Irvine repository is the Post Operative Patient dataset. In which the patient shift from one unit to another after surgery and study of the complete picture of patient's condition to minimize the risk of medical errors and to offer the optimal patient care. The aim is to predict whether a patient will be send to the general hospital floor or dispatch to the home according to their health situation. Moreover, two patients were included in the training data sent to an Intensive care unit. [1]

The paper is arranged as: Section I discusses data mining in medical field. Similar works in field of data mining are presented in Section

DISCUSSION ON PAPERS This section presents with an overview of various research papers contribution related to the techniques of data mining like Multi-layer perceptron, Decision tree, Random forest, Naïve Bayes, K-nearest neighbour for diagnosis in the medical field. Many researchers worked on different datasets in medical field, some of them have been briefly discussed as follows. In 2009 [12], Authors predicted Quasi-Newton method performs well to predict the patient’s post operative patient recovery area by comparing the seven algorithms to train the multi-layered neural network architecture. After that in the next level, Levenberg Marquardt performs better but both these methods are not suitable for large datasets. In 2005 [13], Authors have compared three data mining prediction methods for breast cancer survivability by using 10-fold cross-validation. Decision tree (C5) is predicted as best with accuracy 93.6% on holdout sample, artificial neural network is second with accuracy of 91.2% and logistic regression is found to be third worst with 89.2% accuracy. In 2013 [14], Authors predicted the most efficient model for the patients with lung cancer is Naïve Bayes than decision tree and neural network. In 2011 [15], Authors has predicted that Support Vector Machine found to be best in prediction of cardiovascular disease in patients by comparing RIPPER, decision tree (C4.5), Artificial Neural Network, Support Vector Machine on basis of Accuracy, Sensitivity, True