Long short-term memory (LSTM) neural networks have been proven to be effective for time series prediction, even in some instances where the data is non-stationary. This lead us to examine their predictive ability of stock market returns, as the development of stock prices and returns tend to be a non-stationary time series. We used daily stock trading data to let an LSTM train models at predicting daily returns for 60 stocks from the OMX30 and Nasdaq-100 indices. Subsequently, we measured their accuracy, precision, and recall. The mean accuracy was 49.75 percent, meaning that the observed accuracy was close to the accuracy one would observe by randomly selecting a prediction for each day and lower than the accuracy achieved by blindly predicting all days to be positive. Finally, we concluded that further improvements need to be made for models trained by LSTMs to have any notable predictive ability in the area of stock returns.

The stock market undergoes many fluctuations on a daily basis. These changes can be challenging to anticipate. Understanding such volatility are beneficial to investors as it empowers them to make inform decisions to avoid losses and invest when opportunities are predicted to earn funds. The objective of this research is to use text mining and data mining techniques to discover the relationship between news articles and stock prices fluctuations. There are a variety of sources for news articles, including Bloomberg, Google Finance, Yahoo Finance, Factiva, Thompson Routers, and Twitter. In our research, we use Factive and Intrinio news databases. These databases provide daily analytical articles about the general stock market, as well as daily changes in stock prices. The focus of this research is on understanding the news articles which influence stock prices. We believe that different types of stocks in the market behave differently, and news articles could provide indications on different stock price movements. The goal of this research is to create a framework that uses text mining and data mining algorithms to correlate different types of news articles with stock fluctuations to predict whether to “Buy”, “Sell”, or “Hold” a specific stock. We train Doc2Vec models on 1GB of financial news from Factiva to convert news articles into vectors of 100 dimensions. After preprocessing the data, including labeling and balancing the data, we build five predictive models, namely Neural Networks, SVM, Decision Tree, KNN, and Random Forest to predict stock movements (Buy, Sell, or Hold). We evaluate the performances of the predictive models in terms of accuracy and area under the ROC. We conclude that SVM provides the best performance among the five models to predict the stock movement.

Abstract The impenetrable movement and crash of the stock market is always the most intriguing research task of any financial researcher. Nowadays, it has been proved that the movements of financial asset have the property of non-linearity or near-chaos and shows some tendency within a given period. We used the R/S analysis as the tool to indicate the tendency, and those stocks as our researching objects. We then combined purely price technical analysis indicators and genetic algorithms to form a predicting model. Then we compared our genetic predicting model with the traditional ARIMA analysis and hope to find out the invisible pattern under price volatility. And we hope our model could assist investors in assessing the stock markets more objectively and reduce the risk of stock investment. The researching target is TSMC(2330). We covered the period from 5 September 1994 to 28 December 1999, resulting in 1490 trading days. Historical data are available from Taiwan Economic Journal (TEJ). We execute the researching comparison by bear-market, bull-market, and bull-then-bear market and concluded as follows. 1. After the R/S analysis, we got the Hurst exponent of TSMC to be 0.849855 and the trending cycle was 940. It has proved that the market has tendency and indirectly showed that the Taiwan stock market was not efficient. 2. According to directional precision, our predicting model apparently outpaced the ARIMA model in these three periods. The reason was that our model grabbed more information than the ARIMA model. 3. If we only think about the inputs and outputs, our model seems to be a proper framework for explaining the relationships among variables in comparison with the neural network model having the same input and output variables. 4. We can deduce the invisible relationships of price technical indicators and the closing price. 5. Genetic predicting model can detect the prevailing trend of the learning periods. 6. The shorter the learning period, the better the predicting effects. As a whole and conservatively speaking, we have 70% confidence in directional precision. 7. If we combine proper trading strategy with genetic predicting model and deduct the transaction cost, we still get a better profit than buy-and-hold strategy and have some maneuvering flexibility. 8. After hypothesis testing, our predicting model seems to have some potential of ex ante prediction, but the stability and usability still need further study. In short, we proposed the ex post stock price movement learning model and the viable direction of ex ante prediction. Investors can take advantage of the flexibility of the predicting model and avoid using the over-complex and rigid trading strategies.

abstract: In trading, volume is a measure of how much stock has been exchanged in a given period of time. Since every stock is distinctive and has an alternate measure of shares, volume can be contrasted with historical volume inside a stock to spot changes. It is likewise used to affirm value patterns, breakouts, and spot potential reversals. In my thesis, I hypothesize that the concept of trading volume can be extrapolated to social media (Twitter). The ubiquity of social media, especially Twitter, in financial market has been overly resonant in the past couple of years. With the growth of its (Twitter) usage by news channels, financial experts and pandits, the global economy does seem to hinge on 140 characters. By analyzing the number of tweets hash tagged to a stock, a strong relation can be established between the number of people talking about it, to the trading volume of the stock. In my work, I overt this relation and find a state of the breakout when the volume goes beyond a characterized support or resistance level. / Dissertation/Thesis / Masters Thesis Computer Science 2015

abstract: There have been extensive research in how news and twitter feeds can affect the outcome of a given stock. However, a majority of this research has studied the short term effects of sentiment with a given stock price. Within this research, I studied the long-term effects of a given stock price using fundamental analysis techniques. Within this research, I collected both sentiment data and fundamental data for Apple Inc., Microsoft Corp., and Peabody Energy Corp. Using a neural network algorithm, I found that sentiment does have an effect on the annual growth of these companies but the fundamentals are more relevant when determining overall growth. The stocks which show more consistent growth hold more importance on the previous year’s stock price but companies which have less consistency in their growth showed more reliance on the revenue growth and sentiment on the overall company and CEO. I discuss how I collected my research data and used a multi-layered perceptron to predict a threshold growth of a given stock. The threshold used for this particular research was 10%. I then showed the prediction of this threshold using my perceptron and afterwards, perform an f anova test on my choice of features. The results showed the fundamentals being the better predictor of stock information but fundamentals came in a close second in several cases, proving sentiment does hold an effect over long term growth. / Dissertation/Thesis / Masters Thesis Computer Science 2016

abstract: Stock market news and investing tips are popular topics in Twitter. In this dissertation, first I utilize a 5-year financial news corpus comprising over 50,000 articles collected from the NASDAQ website matching the 30 stock symbols in Dow Jones Index (DJI) to train a directional stock price prediction system based on news content. Next, I proceed to show that information in articles indicated by breaking Tweet volumes leads to a statistically significant boost in the hourly directional prediction accuracies for the DJI stock prices mentioned in these articles. Secondly, I show that using document-level sentiment extraction does not yield a statistically significant boost in the directional predictive accuracies in the presence of other 1-gram keyword features. Thirdly I test the performance of the system on several time-frames and identify the 4 hour time-frame for both the price charts and for Tweet breakout detection as the best time-frame combination. Finally, I develop a set of price momentum based trade exit rules to cut losing trades early and to allow the winning trades run longer. I show that the Tweet volume breakout based trading system with the price momentum based exit rules not only improves the winning accuracy and the return on investment, but it also lowers the maximum drawdown and achieves the highest overall return over maximum drawdown. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2016

This report investigates how prediction of stock markets with Artificial Neural Networks (ANN) is affected by altering aspects of data quantities. A short-term and a long-term perspective considering time delays are examined. Inspired by neurosciences, ANNs have shown great potential in terms of recognising patterns in nonlinear systems. Existing research suggests that ANN is an eminent model to predicting stock markets due to its dynamical characteristics. Closing prices of large-caps within the sectors of IT and Telecommunication represented by the Swedish of OMX30 Stockholm (OMXS30), have been leveraged as data. The ANNs are implemented as multilayer feedforward networks, trained using supervised learning. To identify specific configurations, the models have undergone extensive testing by mean squared errors and statistical analysis. The results obtained suggest that the short-term perspective is optimally predicted for significantly small numbers of time delays, and that optimal configurations do not alter for increasing quantities of data. No significant conclusions could be drawn from the results for the long-term perspective.Key words: ANOVA, Backpropagation, Configurations, Stock Prediction, Artficial Neural Networks

Corporate crisis events such as cyber attacks, executive scandals, facility accidents, fraud, and product recalls can damage customer trust and firm reputation severely, which may lead to tremendous loss in sales and firm equity value. My research aims to integrate information available on the market to assist firms in tackling crisis events, and to provide insight for better decision making. We first study the impact of crisis events on firm performance. We build a hybrid deep learning model that utilizes information from financial news, social media, and historical stock prices to predict firm stock performance during firm crisis events. We develop new methodologies that can extract, select, and represent useful features from textual data. Our hybrid deep learning model achieves 68.8% prediction accuracy for firm stock movements. Furthermore, we explore the underlying mechanisms behind how stakeholders adopt and propagate event information on social media, as well as how this would impact firm stock movements during such events. We adopt an extended epidemiology model, SEIZ, to simulate the information propagation on social media during a crisis. The SEIZ model classifies people into four states (susceptible, exposed, infected, and skeptical). By modeling the propagation of firm-initiated information and user-initiated information on Twitter, we simulate the dynamic process of Twitter stakeholders transforming from one state to another. Based on the modeling results, we quantitatively measure how stakeholders adopt firm crisis information on Twitter over time. We then empirically evaluate the impact of different information adoption processes on firm stock performance. We observe that investors often react very positively when a higher portion of stakeholders adopt the firm-initiated information on Twitter, and negatively when a higher portion of stakeholders adopt user-initiated information. Additionally, we try to identify features that can indicate the firm stock movement during corporate events. We adopt Layer-wised Relevance Propagation (LRP) to extract language features that can be the predictive variables for stock surge and stock plunge. Based on our trained hybrid deep learning model, we generate relevance scores for language features in news titles and tweets, which can indicate the amount of contributions these features made to the final predictions of stock surge and plunge. / Doctor of Philosophy / Corporate crisis events such as cyber attacks, executive scandals, facility accidents, fraud, and product recalls can damage customer trust and firm reputation severely, which may lead to tremendous loss in sales and firm equity value. My research aims to integrate information available on the market to assist firms in tackling crisis events and providing insight for better decision making. We first study the impact of crisis events on firm performance. We investigate five types of crisis events for SandP 500 companies, with 14,982 related news titles and 4.3 million relevant tweets. We build an event-driven hybrid deep learning model that utilizes information from financial news, social media, and historical stock prices to predict firm stock performance during firm crisis events. Furthermore, we explore how stakeholders adopt and propagate event information on social media, as well as how this would impact firm stock movements during the events. Social media has become an increasingly important channel for corporate crisis management. However, little is known on how crisis information propagates on social media. We observe that investors often react very positively when a higher portion of stakeholders adopt the firm-initiated information on Twitter, and negatively when a higher portion of stakeholders adopt user-initiated information. In addition, we find that the language used in the crisis news and social media discussions can have surprising predictive power on the firm stock. Thus, we develop a methodology to identify the importance of text features associated with firm performance during crisis events, such as predictive words or phrases.

Stock market plays an important role in the modern capital market. As a result, the prediction of financial assets attracts people in different areas. Moreover, it is commonly accepted that stock price movement generally follows a major trend. As a result, forecasting the market trend becomes an important mission for a prediction method. Accordingly, we will predict the long term trend rather than the movement of near future or change in a trading day as the target of our predicting approach. Although there are various kinds of analyses for trend prediction, most of them use clear cuts or certain thresholds to classify the trends. Users (or investors) are not informed with the degrees of confidence associated with the recommendation or the trading signal. Therefore, in this research, we would like to study an approach that could offer the confidence of the trend analysis by providing the probabilities of each possible state given its historical data through Dynamic Bayesian Network. We will incorporate the well-known principles of Dow¡¦s Theory to better model the trend of stock movements. Through the results of our experiment, we may say that the financial performance of the proposed model is able to defeat the buy and hold trading strategy when the time scope covers the entire cycle of a trend. It also means that for the long term investors, our approach has high potential to win the excess return. At the same time, the trading frequency and correspondently trading costs can be reduced significantly.

This study investigates the predictive performance of two different machine learning (ML) models on the stock market and compare the results. The chosen models are based on artificial neural networks (ANN) and random forests (RF). The models are trained on two separate data sets and the predictions are made on the next day closing price. The input vectors of the models consist of 6 different financial indicators which are based on the closing prices of the past 5, 10 and 20 days. The performance evaluation are done by analyzing and comparing such values as the root mean squared error (RMSE) and mean average percentage error (MAPE) for the test period. Specific behavior in subsets of the test period is also analyzed to evaluate consistency of the models. The results showed that the ANN model performed better than the RF model as it throughout the test period had lower errors compared to the actual prices and thus overall made more accurate predictions. / Denna studie undersöker hur väl två olika modeller inom maskininlärning (ML) kan förutspå aktiemarknaden och jämför sedan resultaten av dessa. De valda modellerna baseras på artificiella neurala nätverk (ANN) samt random forests (RF). Modellerna tränas upp med två separata datamängder och prognoserna sker på nästföljande dags stängningskurs. Indatan för modellerna består av 6 olika finansiella nyckeltal som är baserade på stängningskursen för de senaste 5, 10 och 20 dagarna. Prestandan utvärderas genom att analysera och jämföra värden som root mean squared error (RMSE) samt mean average percentage error (MAPE) för testperioden. Även specifika trender i delmängder av testperioden undersöks för att utvärdera följdriktigheten av modellerna. Resultaten visade att ANN-modellen presterade bättre än RF-modellen då den sett över hela testperioden visade mindre fel jämfört med de faktiska värdena och gjorde därmed mer träffsäkra prognoser.