Publication: Elaboración de un algoritmo predictivo sobre la evolución del precio de las criptomonedas
Loading...
Official URL
Full text at PDC
Publication Date
2019
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
En este proyecto trataremos de comprobar si una red neuronal es capaz de predecir el precio de una criptomoneda. Esta tarea es complicada porque los precios de las criptomonedas son muy volátiles, lo que supone un reto a la hora de desarrollar la red neuronal.
Comenzaremos por una fase larga de aprendizaje en la que realizaremos cursos de Inteligencia Artificial, tanto básicos como más avanzados. También nos ayudaremos del trabajo previo de otros investigadores y estudiantes.
Ya introducidos en el mundo de la inteligencia artificial, pasaremos a recopilar los datos sobre criptomonedas necesarios para entrenar a nuestra red. Les aplicaremos un preprocesamiento para facilitar el aprendizaje y, una vez estén listos los datos, procederemos al desarrollo de toda la infraestructura necesaria para nuestra red.
Durante este desarrollo nos encontraremos con múltiples complicaciones, como pueden ser la falta de datos o los problemas de formato. Para solucionarlos recopilaremos estos datos de diferentes plataformas y probaremos bibliotecas específicas para el formateo de los mismos.
Con toda la estructura de la red definida pasaremos a una fase de prueba y error en la que comprobaremos cuáles son los factores que más influyen a la hora de realizar una predicción en este ámbito.
Finalmente, expondremos posibles mejoras que podremos llevar a cabo en un futuro si conseguimos los medios necesarios.
El software generado a lo largo del trabajo está disponible en un repositorio público de GitHub. La dirección de este repositorio es la siguiente: https://github.com/PabloDeTorre/NeuralBTCPredictor
In this project we will try to verify if a neuronal network is able to predict the evolution of a cryptocurrency price. This is a complex task because of the volatility of the cryptocurrencies prices, which is a challenge when developing a neuronal network. We will start with a long learning phase in which we will take basic Artificial Intelligence courses and, later on, more advanced ones. We will also help ourselves with the previous work of other researchers and students. Once introduced in the world of Artificial Intelligence, we will collect the data about cryptocurrencies needed to train our network. We will apply a preprocessing to make learning easier. Once the data is ready, we will proceed to develop the entire necessary infrastructure for our network. During this development we will encounter multiple complications, such as be the lack of data or format problems. To solve them we will collect these data from different platforms and we will test libraries specialized in data formatting. With the entire structure of our neuronal network defined, we will go through a trials phase. In this stage, we will check which are the factors that influence the most when making a prediction in this area. Finally, we will discuss possible improvements that we can carry out in the future if we get the necessary resources. All the software created for this project is available in the following GitHub repository: https://github.com/PabloDeTorre/NeuralBTCPredictor
In this project we will try to verify if a neuronal network is able to predict the evolution of a cryptocurrency price. This is a complex task because of the volatility of the cryptocurrencies prices, which is a challenge when developing a neuronal network. We will start with a long learning phase in which we will take basic Artificial Intelligence courses and, later on, more advanced ones. We will also help ourselves with the previous work of other researchers and students. Once introduced in the world of Artificial Intelligence, we will collect the data about cryptocurrencies needed to train our network. We will apply a preprocessing to make learning easier. Once the data is ready, we will proceed to develop the entire necessary infrastructure for our network. During this development we will encounter multiple complications, such as be the lack of data or format problems. To solve them we will collect these data from different platforms and we will test libraries specialized in data formatting. With the entire structure of our neuronal network defined, we will go through a trials phase. In this stage, we will check which are the factors that influence the most when making a prediction in this area. Finally, we will discuss possible improvements that we can carry out in the future if we get the necessary resources. All the software created for this project is available in the following GitHub repository: https://github.com/PabloDeTorre/NeuralBTCPredictor
Description
El software generado a lo largo del trabajo está disponible en un repositorio público de GitHub:
https://github.com/PabloDeTorre/NeuralBTCPredictor
Trabajo de Fin de Grado, Universidad Complutense, Facultad de Informática, Departamento de Sistemas Informáticos y Computación, Curso 2018/2019