DESCRIPTIVE AND PREDICTIVE ANALYSIS OF MAIL ORDER PHARMACY

• Main Author: YOZA PUTRA (NIM : 23516005), HAFID
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27631
• Institution: Institut Teknologi Bandung

<p align="justify">Mail Order Pharmacy (MOP), which is part of online pharmacies that have been implemented in the Americas and Europe. The service initially operates for veteran populations to avoid having to go to a health service center for treatment. MOP has many benefits for consumers including free shipping, 24-hour availability and <br /> <br /> availability of drug stock. In Indonesia there are two services that point to MOP. These services are Mocehat and Halodoc. Data on drug sales transactions in one of the European countries can be described and predict revenue as one of the recommendations of decision-making. <br /> <br /> The transaction data needs to go through the preprocessing stage in order to proceed to the next stage. Description of the business condition of MOP can be done by using data mart and data visualization. Predicted revenue begins with feature selection steps such as correlation and feature deletion. Prediction is done with data mining which consists of two stages: order classification and sale_quantity regression. Classification uses the Naive Bayes method, Multi Layer Perceptron Neural Network and Support Vector Machine. Regression using Multi Layer <br /> <br /> Perceptron Neural Network method. The method is used to build the best model for prediction. The performance measurement of the classification model uses FMeasure <br /> <br /> and regression model using Root Mean Square Error (RMSE). <br /> <br /> The results of data visualization provide information about MOP business condition including the number of non-generic drug sales more than generic drugs. Features <br /> <br /> derived from feature deletion ie rrp, price, competitorprice and pid produce the best classification model. Features gleaned from all features produce the best regression model. The best performance of the classification model test results has F-Measure 0.462 and the regression model has RMSE 1.6654. The predicted results from both models are used for revenue prediction.<p align="justify">