Forex prediction engine: framework, modelling techniques and implementations

Abstract

This paper presents foreign exchange (Forex) prediction engine that included framework, modelling techniques and implementations, to support the needs of financial organizations or individual investors. In the financial sector, Forex prediction is considered to be a complex field, based on the noise of exchange rates as the major challenge. However, there are some financial instruments that are available to guide the individual for future investments. We propose a promising Forex prediction engine using historical Forex data, to extract a pattern movement over a period of time series using Linear Regression Line (LRL) technique and the proposed segmentation algorithm. Subsequently, Artificial Neural Network (ANN) algorithm is applied to classify unique groups of uptrend and downtrend patterns. Dynamic Time Warping (DTW) algorithm is implemented through brute force to identify the similarity trend and the matched result that is used to predict the trend movement for next day. This research reveals the structure of Forex prediction engine by the description of an integrated framework using technical analysis method and machine learning algorithm. The experimental results of AUD -USD and EUR -USD currencies demonstrated 71% of accuracy and its predictions are reliable.

Comparative Study of Forecasting Models for Forex Predictions with the Impact of Different Currencies

“…The methods that are used in stock prediction can also be used for Forex prediction because of the similarity of the market's behavior. [1] proposed a prediction system that uses Linear Regression Line (LRL) for segmentation, ANN for classification, and Dynamic Time Warping (DTW) for recognition. Linear Regression Line is applied for analyzing and forming general trend patterns of historical Forex data of certain currency pairs.…”

“…There are many methods or models that can be used when predicting Forex. In [1] proposed a Forex prediction system that uses Linear Regression Line (LRL), and Dynamic Time Warping (DTW). Linear Regression Line is applied for analyzing and forming general trend patterns of historical Forex data of certain currency pairs.…”

Modelo de análisis multivariante aplicado a la predicción de la tendencia del tipo de cambio euro-dólar

“…En este sentido, la Tabla 8.22 compara la metodología propuesta (LCC) con el estado del arte, y resumen las principales características de las metodología evaluadas. Estas metodología incluyen enfoques ensemble [329] e híbridos [576], redes neuronales [577,578] y predicción de tendencias a partir del análisis de sentimiento [579]. Estos enfoques se han elegido para la comparación no solo por su eficacia, sino también porque son capaces de predecir la dirección del tipo de cambio euro-dólar.…”

“…Por otra parte, la extracción de características es muy común en estos enfoques, lo que compromete la legibilidad de las interpretaciones a cambio de mejoras en la precisión. Aunque los enfoques evaluados han demostrado su poder predictivo tanto con datos de entrenamiento como con datos fuera de muestra (OOS), se debe tener en cuenta que algunos estudios, como [329] y [578], solo se evaluaron con datos dentro de muestra (TVT). Esto puede llevar a una sobreestimación de su capacidad predictiva y a una falta de generalización a nuevos datos.…”