preview

Feature Selection Methods : A Comparison Of Feature Selection Algorithmss

Better Essays
Open Document

For the smallest set containing only ten samples, 19 of the 23 possible feature selection algorithms completed processing (4 feature selection algorithms could not be completed due to the 10-fold cross-validation used). For those 19 feature selection algorithms, 585 classification models were generated (few of the ARFF files were empty for the lower feature thresholds due to the small number of samples). The 50-sample dataset completed 20 of the 23 possible feature selection algorithms, thereby generating 665 classification models. When using 100 samples, 20 of the 23 possible feature selection algorithms were completed, and subsequently utilized to generate 665 classification models. The 200-sample dataset provided 20 of the 23 possible …show more content…

ML classification models were generated using the training data with accuracies of 98.43% for the Tumor vs. Normal model, and 99.06% for the Adenocarcinoma vs. Squamous Cell Carcinoma model. These higher accuracies could also be due to the normalization and background correction performed by the CancerDiscover pipeline. Of all the three compared software tools, GenePattern’s accuracies are most similar to the ones produced by CancerDiscover – 99.21% and 99.06%, respectively. All probes were utilized in the model building since feature selection could not be performed using GenePattern. On the other hand, CancerDiscover was able to achieve similar accuracy by using as few as three probes (See Table 3). Finally, CancerDiscover differs from the proprietary GenePattern by the fact that CancerDiscover is open-source; as such, its methodologies are transparent and reproducible, and the community can further expand the software.

Chipster is developed based on a client-server architecture. Data is imported at the client side, while all processing is performed on the server side using R programming. It requires that all data needs to be transferred between client and server for each analysis step which can be very time-consuming if the datasets are large (Koschmieder et al., 2012). Chipster was not able to successfully perform a classification when we provided the dataset containing all probes. As a result, feature

Get Access
Related
Get Access