Impact of Data Size on Data Mining Prediction Accuracy

Prediction accuracy is an essential topic to be researched and studied to have maximum precision and more efficient prediction systems especially in critical fields like medicine and bioinformatics. In this study we investigated the prediction models, viz., Logistic Regression, Neural Network and Decision Tree and compared their prediction performance taking into account the dataset size and the dependent variable values distribution in the dataset as prediction model performance metrics and we used a robust and highly correlated data from the Palestinian Households Census 2007. The Neural Network outperformed the other two models and the Logistic Regression outperformed the Decision Tree. The dependent variable values distribution and dataset size had no significant effect on prediction accuracy but the data content was found to as the most affecting factor on prediction accuracy. Depending on the results of this research we believed and stated that the prediction accuracy is a function of data content and characteristics.