Colombian Coffee Price Forecast via LSTM Neural Networks
This work deals with the contributions Machine Learning techniques can bring into the coffee growing conglomerate, committees and other points in the production and marketing chain involved in the dynamics of this commodity. It is well known that the different variables that interact with prices bot...
- Autores:
-
Herrera Jaramillo, Yoe Alexander
Ortega Giraldo, Johana C.
Acevedo Amorocho, Alejandro
Prada Marín, Duwamg Alexis
- Tipo de recurso:
- Part of book
- Fecha de publicación:
- 2021
- Institución:
- Tecnológico de Antioquia
- Repositorio:
- Repositorio Tdea
- Idioma:
- eng
- OAI Identifier:
- oai:dspace.tdea.edu.co:tdea/3958
- Acceso en línea:
- https://dspace.tdea.edu.co/handle/tdea/3958
- Palabra clave:
- Aprendizado de máquina
Café
Coffee
Precios
Prix
Prices
Preço
Machine learning
Aprendizaje automático
Apprentissage machine
Neural Networks, Computer
Redes Neurais de Computação
Redes Neurales de la Computación
- Rights
- closedAccess
- License
- http://purl.org/coar/access_right/c_14cb
| Summary: | This work deals with the contributions Machine Learning techniques can bring into the coffee growing conglomerate, committees and other points in the production and marketing chain involved in the dynamics of this commodity. It is well known that the different variables that interact with prices both nationally and internationally have a direct, dramatic affect on the sector under study. In this work, we summarize an extensive review of the coffee price dynamics and the forecast techniques used in this eld. In addition, the internal coffee price in Colombia has been modeled using a long short-term memory (LSTM) recurrent neural network that was chosen as the one of better performance out of three original models. The archetype that evidenced a pertinent superiority of fitness within the parameters specified for this type of model is composed of a linear self-regressive component, plus a multi-layer perceptron-type artificial neural network with twenty (40) LSTM cells neurons in the hidden layer. This epitome captures the chaotic coffee price dynamics. The normalized residuals of the model are uncorrelated and homoscedastic and follow a normal distribution. The results indicate that the current price depends on the prices that occurred in the last four (4) years. This tool can be used to help the coffee growing community to better design alternatives to overcome difficulties with the price of the grain, and this makes it a Logistics solution for them. |
|---|