Selection of most informative features has found to be a great utility in minimising the dimensions of high dimensional data and thereby on increasing the learning efficiency of learning algorithms. The majority of attribute selection methods ignored the consideration of feature multi-correlation such as feature complementarity and interaction correlation. Also there is less research on influence of feature selection algorithms on performance improvement of machine learning algorithms for cervical cancer risk factor dataset. Motivated by this, we performed experiments by applying information theoretic based feature selection algorithms that considers class based feature multicorrelation on cervical cancer dataset. We considered eight different machine learning algorithms (MLA) for performance comparison and analysis in our research. The feature selection algorithms considered for this research are DISR, R2CI, DRJMIM, DWFS, IWFS and UCRFS. The metrics used for MLA comparison are accuracy, precision, recall, confusion matrix and auc. The findings illustrate the significance of DRJMIM feature classification technique and RandomForest classifier for cervical cancer prediction. RandomForest classifier shows highest classification accuracy of 98.55% after applying DRJMIM FSA.