Andrés Arévalo

High-Frequency Trading Strategy Based on Deep Neural Networks

This paper presents a high-frequency strategy based on Deep Neural Networks (DNNs). The DNN was trained on current time (hour and minute), and \( n \)-lagged one-minute pseudo-returns, price standard deviations and trend indicators in order to forecast the next one-minute average price. The DNN predictions are used to build a high-frequency trading strategy that buys (sells) when the next predicted average price is above (below) the last closing price. The data used for training and testing are the AAPL tick-by-tick transactions from September to November of 2008. The best-found DNN has a 66 % of directional accuracy. This strategy yields an 81 % successful trades during testing period.

Deep Learning and Wavelets for High-Frequency Price Forecasting

This paper presents improvements in financial time series prediction using a Deep Neural Network (DNN) in conjunction with a Discrete Wavelet Transform (DWT). When comparing our model to other three alternatives, including ARIMA and other deep learning topologies, ours has a better performance. All of the experiments were conducted on High-Frequency Data (HFD). Given the fact that DWT decomposes signals in terms of frequency and time, we expect this transformation will make a better representation of the sequential behavior of high-frequency data. The input data for every experiment consists of 27 variables: The last 3 one-minute pseudo-log-returns and last 3 one-minute compressed tick-by-tick wavelet vectors, each vector is a product of compressing the tick-by-tick transactions inside a particular minute using a DWT with length 8. Furthermore, the DNN predicts the next one-minute pseudo-log-return that can be transformed into the next predicted one-minute average price. For testing purposes, we use tick-by-tick data of 19 companies in the Dow Jones Industrial Average Index (DJIA), from January 2015 to July 2017. The proposed DNN’s Directional Accuracy (DA) presents a remarkable forecasting performance ranging from 64% to 72%.

Clustering algorithms for Risk-Adjusted Portfolio Construction

Abstract This paper presents the performance of seven portfolios created using clustering analysis techniques to sort out assets into categories and then applying classical optimization inside every cluster to select best assets inside each asset category. The proposed clustering algorithms are tested constructing portfolios and measuring their performances over a two month dataset of 1-minute asset returns from a sample of 175 assets of the Russell 1000® index. A three-week sliding window is used for model calibration, leaving an out of sample period of five weeks for testing. Model calibration is done weekly. Three different rebalancing periods are tested: every 1, 2 and 4 hours. The results show that all clustering algorithms produce more stable portfolios with similar volatility. In this sense, the portfolios volatilities generated by the clustering algorithms are smaller when compare to the portfolio obtained using classical Mean-Variance Optimization (MVO) over all the dataset. Hierarchical clustering algorithms achieve the best financial performance obtaining an adequate trade-off between accumulated financial returns and the risk-adjusted measure, Omega Ratio, during the out of sample testing period.

RFID framework for intelligent traffic monitoring

Nowadays the transportation systems show new challenges in management and supervision issues due to amount of vehicles and the variability and extension of the roads. New technologies have been developed to get information about traffic in cities. Nevertheless, new frameworks with different technologies applied to traffic monitoring have not been well defined. In this paper, a new framework with Radio Frequency Identification technology (RFID) applied to road traffic monitoring is exposed. In order to validate the designed framework, an experiment in a university campus was carried out, concluding the effectiveness of the RFID technology for traffic monitoring.

CNN with Limit Order Book Data for Stock Price Prediction

This work presents a remarkable and innovative short-term forecasting method for Financial Time Series (FTS). Most of the approaches for FTS modeling work directly with prices, given the fact that transaction data is more reachable and more widely available. For this particular work, we will be using the Limit Order Book (LOB) data, which registers all trade intentions from market participants. As a result, there is more enriched data to make better predictions. We will be using Deep Convolutional Neural Networks (CNN), which are good at pattern recognition on images. In order to accomplish the proposed task we will make an image-like representation of LOB and transaction data, which will feed up into the CNN, therefore it can recognize hidden patterns to classify FTS in short-term periods. We will present step by step methodology to encode financial time series into an image-like representation. Results present an impressive performance, ranging between 63% and 66% in Directional Accuracy (DA), having advantages in reducing model parameters as well as to make inputs time invariant.

Price Prediction with CNN and Limit Order Book Data

This work introduces how to use Limit Order Book Data (LOB) and transaction data for short-term forecasting of stock prices. LOB registers all trade intentions from market participants, as a result, it contains more market information that could enhance predictions. We will be using Deep Convolutional Neural Networks (CNN), which are good at pattern recognition on images. In order to accomplish the proposed task we will make an image-like representation of LOB and transaction data, which will feed up into the CNN, therefore it can recognize hidden patterns to classify Financial Time Series (FTS) in short-term periods. Data enclose information from 11 NYSE instruments, including stocks, ETF and ADR. We will present step by step methodology for encoding financial time series into an image-like representation. Results present an impressive performance, 74.15% in Directional Accuracy (DA).

Algorithmic Trading Using Deep Neural Networks on High Frequency Data

In this work, a high-frequency trading strategy using Deep Neural Networks (DNNs) is presented. The input information consists of: (i). Current time (hour and minute); (ii). The last n one-minute pseudo-returns, where n is the sliding window size parameter; (iii). The last n one-minute standard deviations of the price; (iv). The last n trend indicator, computed as the slope of the linear model fitted using the transaction prices inside a particular minute. The DNN predicts the next one-minute pseudo-return, this output is later transformed to obtain a the next predicted one-minute average price. This price is used to build a high-frequency trading strategy that buys (sells) when the next predicted average price is above (below) the last closing price.