RESEARCH ON THE PRINCIPLES OF BUILDING AND DESIGNING AN INTERACTIVE PLATFORM USING DATA PREDICTION METHODS AND ALGORITHMS

Abstract

Automation of processes is one of the main directions in the development of modern technologies. It significantly increases efficiency and reduces risks by ensuring the speed and quality of task execution of various types. This article presents a study on the development of a rental platform for various items that includes a mechanism for predicting the probability of disputes. The platform is built using a multi-server architecture with a clear application of specific technologies to solve particular tasks. Special attention is given to the review and analysis of possible models for predicting the probability of disputes. Several machine learning methods were considered, namely logistic regression combined with prior text-to-vector transformation, and the random forest algorithm, as this method can work with diverse sets of numerical data. The potential use of neural networks was also analyzed, as they are capable of detecting complex nonlinear dependencies in data, which is beneficial for this type of project. As a result, it was decided to use the random forest algorithm as the main model for training data prediction, since it does not require large amounts of data or significant computational resources compared to neural networks. In addition, the project includes functionality that allows administrators to create custom models based on selected fields. Moreover, to avoid unexpected results during training, the platform includes a recommendation mechanism. It uses the χ² test to identify dependencies between categorical variables and a correlation matrix to determine linear relationships between numerical data, allowing the selection of the most relevant fields. The study also presents the structure of the project’s database and describes the use of key table fields to improve training accuracy. As a result, a platform was created that simplifies the identification of the most suitable opponent for initiating an order.