Exploring the Impact of Filtering Techniques on the Performance of Different Classifiers for Predicting Diabetes Mellitus

Diabetes is a chronic disorder that affects how
many people around the world process their metabolism. The
prevalence of diabetes is increasing alarmingly every year.
Diabetes can harm various essential organs in the body if it is
not managed well. Therefore, it is very important to identify
diabetes early and initiate treatment as soon as possible to avoid
the disease from causing such complications. This research
applied five different techniques in WEKA tools to forecast
diabetes based on the input attributes of the dataset. This
research used 17 attributes, such as Age, Sex, Polyuria,
Polydipsia and other medical terms to assess the likelihood of a
patient developing disease. Nowadays Filtering techniques are
crucial in various fields and applications, serving to extract
relevant information, enhance data quality, and improve overall
system performance. The filters—Resample, Obfuscate, and
Discretizeare important for preprocessing data for machine
learning algorithms. They help to overcome challenges related
to data imbalance, privacy protection, and data representation,
resulting in enhanced model performance and data usability.
These filters are vital for achieving high-quality data
preprocessing, feature engineering, and privacy preservation in
machine learning workflows. Preprocessed data applied to the
classification techniques, namely JRip, SMO, KNN, Logistic and
SGD used to examine the diabetes. The performance
measurement for this study are the mean absolute error and
kappa statistics and the accuracy of correct classification, of the
classifier. The results show that Overall, employing these filters
in the preprocessing stage assists to optimize data preparation
and establishes the basis for successful machine learning
outcomes