Performance Evaluation of Classification Models

As the amount of data is increasing over the tremendous rate, it is extremely viable to imply smart analysis. It deals with optimization of performance criterion dealing with examples relevant to present and past situations. Learning plays a vital role in making predictions from analysis of data set properties. We also ensure that the selected data set suits our purpose of predicting futuristic events and unseen samples. However while dealing with problems of classification in machine learning we need to determine and draw observations relevant to a problem statement having disjoint set of training data. Mining information from data enrols classification, clustering and other such methodologies as its subsets. The thesis presents a classical descriptive procedure to compare various classification schemes under single roof and draw analysis over the best scorer in terms of accuracy to draw predictions of Bank Marketing success rate while targeting potential customers. Various datasets can be filtered by the approached schemes to make decision during binary and multi valued classification. However the classification model ranks one over other to determine the best fit choice.