Type: Process Essays
Sample donated: Stephen Nunez
Last updated: September 18, 2019
AbstractIn today’s world Breastcancer is one of the major problem faced by women . Identifying cancer is theprimitive stage and is stillchallenging.
The diagnosis and treatment of the breast cancer have become anurgent. Breast cancer, is widely seen tumor in Indian women . Early treatment of breast cancer have become anextremely crucial work to do, not onlyhelps to cure cancer but also helps in curative of its occurence. Today , there are different kinds of methods and data mining techniquesand various process like knowledge discovery are developed for predicting the breast cancer. As per the study , we perform a comparison ofdifferent classification and clustering algorithms. Various classification algorithms and theclustering algorithm are used.
The result indicate that the classificationalgorithms are better predictors than the clustering algorithms. IntroductionNow-a-daysbreast cancer is common in women. Prognosis of breast cancer is as important asits treatment. Breast cancer is the most common cause of death among women. Ifbreast cancer predicted at its earlier stages,better treatment can be providedwhich enable the person to survive.Diagnosis and treatment of breast cancer hasbecome an urgent work to perform.Different datamining methods are used toretrieve valuable information from large databases inorder to make decisions toprovide better health services.
Breast cancer begins withthe abnormal growth of some breast cells. These cells divide more rapidly andcontinue to accumulate than healthy cells do, forming a lump or mass. These cellsmay grow through your breast to your lymph nodes or to other parts of your body.Breast cancer varies on the basis ofage groups, it is less common at a young age (i.e., in their thirties), youngerwomen lean to have more aggressive breast cancers than older women.
In this paper we performcomparison on different classification as well as clustering algorithm topredict breast cancer. A number of attributes are used in performing comparison.These attributes are compared to find the best classification algorithm.Literature surveyIn paper 1, three different data miningclassification methods are used for the prediction of breast cancer. Itconsiders different parameters for prediction of cancer. But for superiorprediction, focus is on accuracy and lowest computing time.
Studies filteredall algorithms based on lowest computing time and accuracy and it came up with the conclusion that Naïve Bayesis a superior algorithm compared to decision tree and k-nearest neighbor,because it takes lowest time i.e. 0.02 seconds and at the same time isproviding highest accuracy.
In 2 paper,WPBC dataset is used for finding an efficient predictor algorithm to predictthe recurring or non-recurring nature of disease. This helps Oncologists todifferentiate a good prognosis (non-recurrent) from a bad one (recurrent) andcan treat the patients more effectively. Eight popular data mining methods havebeen used, four from clustering algorithms (Kmeans ,EM, PAM and Fuzzy c-means)and four from classification algorithms (SVM, C5.
0, KNN and Naive Bayes).Theresults of these algorithms are clearly outlined in this paper with necessaryresults. The classification algorithms, C5.0 and SVM have shown 81% accuracy inclassifying there occurrence of the disease. This is found to be best amongall. On the other hand, EM was found to be the most promising clusteringalgorithm with the accuracy of 68%. The research shows that the classificationalgorithms are better predictor than clustering algorithms. The impact factorsof various parameters responsible for predicting the occurrence/non-occurrenceof the disease can be verified clinically.
Further, the identified criticalparametersshould be verified by applying on larger medical dataset topredictthe recurrence of the disease in future.In paper 3, they intend to build a diagnosticmodel for breast cancer which is to search the relationship between breastcancer and its symptoms. A feature selection method, INTERACT, is applied toselect related and important features in order to improvethe accuracy of the diagnostic model.
And, SVM is applied to build theclassification model. Two diagnostic models are built with and without featureselection for the sake of proving the significance of the feature selection.Through the experiments, the accuracy of the diagnostic model with featureselection is improved obviously compared with the model without featureselection. Meantime, nine features are chosen out as the relevant factors forbuilding the diagnostic model.
The information found out in this study can besupplementary information for related practitioner better diagnosing heartdisease.In paper 4itfocus on the importance of feature selection in breast cancer prognosis. Usingproper attribute selection technique, any classification algorithm can beimproved significantly. Attributes with less contribution in dataset oftenmisguides the classification and results in poor prediction. In this work, theyfound Support Vector Machine giving much better output both before and afterattribute selection.
Area under ROC curve analysis showed results in favor,where Naïve Bayes and Decision Tree showed much better improvement afterfeature selection method. In this paper we only focused on whether breastcancer is recursive or not. In addition of this work, they try to predict thetime of recurrence of cancer which is classified as recursive.
Paper 5 presented a survey ofclassification simulations which can be used for breast cancer detection using WEKAtool. A discussion on a variety of classification techniques that already existin real world and the performance accuracy is listed from that. By using thatwe can decide which algorithm is best for the WEKA tool for breast cancerdetection.
It compares different algorithmsand found SVM is better having highaccuracy and expectation maximization with the least accuracy.In paper 6 paperpresented a survey of classification simulations which can be used for breastcancer detection using WEKA tool. A variety of classification techniques thatalready exist in real world are discussed. By using that we can decide whichalgorithm is best for the WEKA tool for breast cancer detection. Classification Algorithms Clustering Algorithms Algorithms Confusion Matrix Accuracy Algorithms Confusion Matrix Accuracy C5.0 N R N 47 0 R 11 0 0.
8103 K-Means N R N 100 48 R 23 23 0.6340 KNN N R N 47 0 R 11 0 0.7068 EM N R N 117 31 R 31 15 0.6804 Naïve Bayes N R N 47 0 R 11 0 0.5344 PAM N R N 64 84 R 29 17 0.4175 SVM N R N 47 0 R 11 0 0.8103 Fuzzy c-Means N R N 50 98 R 24 22 0.
3711 Table :comparison of clustering and classificationalgorithms2 Accuracy=(TP+TN)/(TP+TN+FP+FN)TP: True PositiveTN: True NegativeFP: False PositiveFN: False Negative Conclusion From the abovecomparisons we came up with a conclusion that the classification algorithmsworks better than the clustering algorithms in predicting breast cancer. Andinthe classification algorithms the SVM and C5.0 came up with better performance.The best algorithm for predicting breast cancer is purely based on the accuracyof the algorithm.
Reference1 ChintanShah; Anjali G.Jivani “Comparison of data mining classificationalgorithms for breast cancer prediction”2 Uma Ojha; Savita Goel “A study on prediction of breast cancer recurrence using data mining techniques” 2017 7th International Conference on Cloud Computing,Data Science & Engineering – Confluence3 Runjie ShenYuanyuan YangFengfeng Shao “Intelligent Breast Cancer Prediction Model Using Data MiningTechniques”4 Ahmed Iqbal Pritom; Md. Ahadur Rahman Munshi; ShahedAnzarusSabab;Shihabuzzaman Shihab.”Predicting breast cancer recurrence using effective classification and feature selection technique”5S.Padmapriya, M.Devika,V.Meena,S.B.Dheebikaa.Vinodhini , ” Surveyon Breast Cancer Detection Using Weka Tool”6 Jahanvi Joshi, RinalDoshi, Jigar Patel, Ph.D,” Diagnosis ofBreast Cancer using Clustering Data Mining Approach”