A comparative analysis of data normalization on data mining classification performance

Data are a collection of information in the form of facts. Information is stored in data from various origins. Data processing is an important step that is currently carried out. Data processing is commonly performed using data mining. However, data processing usually face barriers that keep it from fully running well because the data stored in the dataset sometimes are not in a normal form. One of the problems encountered in random data is that there is a considerable distance between data, which sets an impediment to data processing. This problem can be solved using normalization. Normalization is also generally referred to as simplification. Some algorithms such as the min-max normalization and Z-score algorithms can be used for normalization. The results of the testing on the use of the min-max normalization and Z-score algorithms for normalization revealed that the former had better performance than the latter. This was judged from the magnitude of the increase in accuracy obtained from the use of both algorithms, in which case min-max normalization gained an increase of 0.41%, while Z-score normalization did an increase of 0.14%.