The Role of Machine Learning in Productivity: A Case Study of Wine Quality Prediction

Türkçe

Artificial intelligence has been used in many areas in recent years and has achieved quite successful results. Like using artificial intelligence from healthcare to driverless vehicles, it also has often been used to increase productivity in the production sector. In this study, we tried to draw a framework for the use of artificial intelligence algorithms in a data set that is not normally distributed. Any artificial intelligence algorithm can be easily applied on normally distributed data sets, while data sets that do not normally distributed require a different operation to the data itself or it is necessary to revise the theoretical structure of the algorithm. In this regard, three different methodologies are applied in this study. Initially, Support Vector Machines, which are often used in the literature, is used. In addition, Weighted Support Vector Machines, which is the revised version of the Support Vector Machines to produce successful results in abnormal distributed data sets. Finally, the Synthetic Minority Oversampling Technique (SMOTE) is applied and the data set used was artificially converted to normal distribution. Three techniques are compared in terms of sensitivity, specificity, precision, prevalence, F-1 score, and G-Mean evaluation criteria were compared. According to the results of the study, Weighted Support Vector Machines produced the most successful results according to the evaluation criteria used.