Universitas Indonesia Conferences, The 4th International Conference of Vocational Higher Education

Font Size: 
Stock Price Forecasting using Adaptive Neural Fuzzy Inference System (ANFIS)
Yulius Eka Agung Seputra

Last modified: 2019-09-19


Manuscript type : Research paper

Research aims : This study aims to predict stock prices by using variables open, high, low, close using artificial neural networks, especially adaptive neural fuzzy inference systems (ANFIS)

Design/  Methodology/ Approach : The survey was conducted to collect stock data from Yahoo Finance sites. The stock data used are data from 2001-2018. Learning patterns of data patterns use Adaptive Neural Fuzzy Inference System (ANFIS) compared to regression analysis, Mean Square Error (MSE) and Mean Prediction Error.

Research findings: The results show that stock price predictions using Adaptive Neural Fuzzy Inference System (ANFIS) have a small error rate (below 1 percent). The stock price when closing (close) is determined by the open price and the volume of the stock. The value of the highest (high) price of the stock and the lowest value of the stock follow the determined value of the opening price (open).

Theotical contributions / Originality: This paper contributes to existing research in the economic field, especially stock investment and Financial Technology. Adaptive neural fuzzy inference system (ANFIS) approach has the advantage of being able to recognize patterns of random patterns that are not recognized when using regression and statistical approaches. Mean Square Error (MSE), Mean Prediction Error (MPE) is used as a model indicator so that predictions do not occur.

Practitioner/ Policy implications : The findings of this study indicate that in predicting stock prices is highly dependent on the opening price and the volume of shares. Each stock has a different and predictable pattern if it has complete data. Stock data providers in Indonesia need to provide complete stock data if they want to implement this research.

Research limitation: This research is limited by stock data for 2001-2018. Therefore, it is very possible to develop using another Artificial Neural Network approach such as Genetic Algorithm (GA)

Full Text: Untitled