Comparison of Four Data Mining Algorithms for Predicting Colorectal Cancer Risk

 Background and Objective: Colorectal cancer (CRC) is one of the most prevalent malignancies in the world. The early detection of CRC is not only a simple process, but it is also the key to its treatment. Given that data mining algorithms could be potentially useful in cancer prognosis, diagnosis, and treatment, the main focus of this study is to measure the performance of some data mining classifier algorithms in terms of predicting CRC and providing an early warning to the high-risk groups.  Materials and Methods: This study was performed in ۴۶۸ subjects (۱۹۴ CRC patients and ۲۷۴ non-CRC cases). We used the CRC dataset from the Imam Hospital, Sari, Iran. The Chi-square feature selection method was utilized to analyze the risk factors. Then, four popular data mining algorithms were compared based on their performance in predicting CRC, and, finally, the best algorithm was identified.  Results: The best outcome was obtained by J-۴۸ (F-Measure = ۰.۸۲۶, ROC=۰.۸۸۱, precision= ۰.۸۲۶ and sensitivity =۰.۸۲۷), Bayesian Net was the second-best performer (F-Measure = ۰.۷۱۸, ROC=۰.۷۸۴, precision= ۰.۷۱۹ and sensitivity=۰.۷۲۲). Random-Forest performed the third-best (F-Measure= ۰.۷۰۵, ROC=۰.۷۵۸, precision= ۰.۷۱۹, and sensitivity=۰.۷۱۲). Finally, the MLP technique performed the worst (F-Measure = ۰.۷۰۲, ROC=۰.۷۶, precision = ۰.۷۰۱ and sensitivity=۰.۷۰۳).                                                                        Conclusion: According to the results, we concluded that the J-۴۸ could provide better insights than other proposed prediction models for clinical applications.

Keywords:

Data Mining , classification models , Colorectal Cancer , prediction