Hybrid ARIMA/ANN vs naive model for trading

2018-05-20T04:48:37Z (GMT) by Marco Repetto

Abstract
This paper presents the combination of the ARIMA
process with ANN. In particular, ARIMA is used to
capture the underlying linearities of stock movement
whereas the error terms are fed in an ANN given as
inputs the price of the stocks from the same market.
Such hybrid ARIMA is used for signal generation in
the prediction of stock prices movement, especially in
such case the goodness of the model is tested in com-
parison with a naive algorithm that generates signals
at random. The portfolio of such game is built using
common stocks from NASDAQ and an intraday time
series of 90 intervals of one minute each leveraging
Figure 1: Percentage of order generated by algo-
the Alphavantage API.