SKU/Artículo: AMZ-B09P1VTHWT

Financial Time Series Forecasting using Neural Networks

Detalles del producto
Disponibilidad:
En stock
Peso con empaque:
0.24 kg
Devolución:
Condición
Nuevo
Producto de:
Amazon
Viaja desde
USA

Sobre este producto
  • Aim of this paper is to explore the application of Artificial Neural Networks (ANNs) for financial time series forecasting. In this paper, a Long Short-Term Memory network (LSTM), which is a special kind of recurrent neural network (RNN), was used for research purposes. LSTMs can deal with the vanishing gradient problem that plague regular RNNs and have the ability to build a long-term memory of previously seen data points. The research objective is to predict the daily percentage returns of the DAX for an observation period of 1.400 days from the 20th of May 2014 until the 21st of November 2019. To facilitate this, four different datasets including macroeconomic and consumer sentiment data were used. The best performance was achieved using basic index data together with macroeconomic factors, resulting in an overall mean directional accuracy (MDA) of 0.5275. The fully trained LSTM was then evaluated using statistical metrics. Compared to a simple naïve forecast, the LSTM model performs worse in every aspect, except for MDA. To test practicability of said model, a Buy & Sell trading strategy is derived and benchmarked against a simple Buy & Hold (B&H) strategy. The LSTM strategy generates higher returns compared to B&H or naïve strategies. Transaction cost destroy most of the profits, however, and before implementing such a trading strategy in the real world, several factors such as the “Lucky Event Issue” or persistence/self-destruction of profits need to be considered
AR$19.424
49% OFF
AR$9.961

IMPORT EASILY

By purchasing this product you can deduct VAT with your RUT number

AR$19.424
49% OFF
AR$9.961

Pagá fácil y rápido con Mercado Pago o MODO

Llega en 8 a 12 días hábiles
con envío
Tienes garantía de entrega
Este producto viaja de USA a tus manos en