Abstract

In this paper, we study the performance of various models for automated evaluation of descriptive answers by using rank based feature selection filters for dimensionality reduction. We quantitatively analyze the best feature selection technique from amongst the five rank based feature selection techniques, namely Chi squared filter, Information gain filter, Gain ratio filter, Relief filter and Symmetrical uncertainty filter. We use Sequential Minimal Optimization with Polynomial kernel to build models and we evaluate the models across various parameters such as Accuracy, Time to build models, Kappa, Mean Absolute Error and Root Mean Squared Error. Except with Relief filter, for all other filters applied models, the accuracies obtained are at least 4% better than accuracies obtained with models with no filters applied. The accuracies recorded are same across Chi squared filter, Information gain filter, Gain ratio filter and Symmetrical Uncertainty filter. Therefore accuracy alone is not the determinant in selecting the best filter. The time taken to build models, Kappa, Mean absolute error and Root Mean Squared Error played a major role in determining the effectiveness of the filters. The overall rank aggregation metric of Symmetrical uncertainty filter is 45 and this is better by 1 rank than the rank aggregation metric of information gain attribute evaluation filter, the nearest contender to Symmetric attribute evaluation filter. Symmetric uncertainty rank aggregation metric is better by 3, 6, 112 ranks respectively when compared to rank aggregation metrics of Chi squared filter, Gain ratio filter and Relief filters. Through these quantitative measurements, we conclude that Symmetrical uncertainty attribute evaluation is the overall best performing rank based feature selection algorithm applicable for auto evaluation of descriptive answers.