Predictive Analysis of Lung Cancer Recurrence

Abstract

Abstract. The paper is about the predictive analysis of lung cancer recurrence based on non-small cell lung cancer carcinoma gene expression data using data mining and machine learning techniques. Prediction is one of the most significant factors in statistical analysis. Predictive analysis is a term describing a variety of statistical and analytical techniques used to develop models that predict future events or behaviours. Prediction of cancer recurrence has been a challenging problem for many researchers. The proposed method involves four phases: data collection, gene selection, designing classifier model, statistical parameter calculation and finally the comparison with previous results. The major part of the method is the gene selection and classification. A hybrid method for gene selection and classification is used for statistical analysis of lung cancer recurrence. The most suitable techniques are used for this work on the basis of comparative analysis of different classification method and optimization techniques.

Cancer adjuvant chemotherapy prediction model for non‐small cell lung cancer

“…Win et al [19] applied naive Bayes (NB) to predict cancer recurrence over three datasets. Srivastava et al [20] also used NB to predict lung cancer recurrence. Kawata et al [21] used a decision tree (DT) algorithm to classify NSCLC data and predict recurrence.…”

“…The NB‐based classification model is easy to implement, but it becomes complicated when there are many attributes in the datasets. The main problem for DT approach is the complicated training process when DT is applied on datasets that contain missing data [20]. These problems in the traditional machine learning approaches make them unsuitable for dealing with complicated data such as microarray data.…”

Prediction of NSCLC recurrence from microarray data with GEP

“…They concluded that it was possible to classify the cancer recurrence successfully by combining the results generated from their proposed gene network models. Win et al [15] and Srivastava et al [16] applied Naïve Bayes (NB) to predict the lung cancer recurrence. Kawata et al [17] obtained a decision tree (DT) for five‐category classification to detect the free survival for NSCLC recurrence.…”

“…Using NB for classification is very easy to design and implement but, like the neural network, many parameters could mislead the classification. For the DT approach, it has two main problems: one is the high complexity when there are errors in the training sets and another is the difficulty of making correct decisions when there are missing values in the datasets [16].…”