Bilgisayar Bilimleri

Yazarlar: Emrah AYDEMİR, Feyzi KAYSİ, Mehmet YAVUZ

Konular:-

Anahtar Kelimeler:Drug Prediction,Tree Algorithms,Data Mining

Özet: Drugs, which play a key role in improving people's deteriorating health balance, are now increasingly being purchased. It is important to know the buying behavior of these drugs. The aim of this study is to predict subsequent situations using data for drug purchase in a region. In this way, the storage status of drugs can be controlled through the types of drugs consumed locally. In this study one year 108525 Total pharmaceutical sales data between a pharmacy in Turkey in March 2019 with the month of May 2018 were analyzed. In this context, sales amounts were tried to be estimated according to patient types, drug writing institution and drug group. In the estimations made with the Weka program, the tree algorithms DecisionStump, M5P, RandomForest, RandomTree and REPTree were used. The mean absolute error, root mean squared error and correlation coefficient values ​​of these algorithms were compared to find the most successful estimation model. While the purchase amount was estimated according to the drug group and the purchase amounts according to the patient types, the correlation coefficients of 0,53 and 0,58 were very low and a successful estimate was not obtained. However, high successful estimates were obtained with a correlation coefficient of 0,83 obtained from the drug issuing institution.

---