Análisis de las Alternativas del Diseño de un Sistema Difuso para la Predicción del Precio del Bitcoin (BTC)

Abstract

Objective: To present three alternatives for the design of a diffuse system that has, as a main objective, the forecast of the Bitcoin (BTC) Price in USA dollars.\xa0\xa0 Methodology:\xa0 To perform a mathematical analysis and graphs of Bitcoin prices [BTC]. The entry prices have been selected as the closing prices for each period, where the period is a constant of 30 minutes, seeking to identify possible characteristics of interest, periodic patterns, and/or the presence of noise.\xa0 Results:\xa0 In the first approach to the problem’s solution, a diffuse system is built with a Mamdani motor of inference where the machine parameters are manually adjusted, based on the patterns found in the previous study of the signal, and in the knowledge of an expert in the area.\xa0 The second solution is made by using an ANFIS algorithm that carries out an automatic adjustment of parameters using diminishing gradient algorithms, and the third solution is designed through ˜ the implementation of a bio-inspired algorithm known as a simple genetic algorithm. Conclusions:\xa0 In all three cases, the entries to the system are five, and correspond to price signals and the closing is the price forecast for the following period.