Home

About us

Methodologies

Progress

Results

References

Updated on 17/04/2019

Objective To test the effectiveness of some technical indicators For the classifier to be developed: 1. Programming language used: Python. There are Python libraries available for easier implementation of machine learning algorithms Weka. There are Python libraries available for easier implementation of machine learning algorithms 2. Possible algorithms used: Many. For example, Decision Tree is going to be explored as the algorithm can show how a prediction is made. This can be instrumental in determination of which indicators are more effective in making correct prediction, which is an objective of this project. Some of the algorithms may be combined to form an ensemble classifier if their accuracy is higher than the others. For the data: 1. Stock quotes 2. Indicators A. Financial indicators They reflect the profitability and cost-effectiveness of a company. For example, there is an indicator called earnings per share (EPS) which measures the profit a company can get for each share. It is expected that indicators like EPS are part of the input because they are frequently reported in financial news. People may decide if they are going to invest in a stock based on what they have known from the financial news. B. Technical indicators They are calculated from historic market data. They may provide one better insight into the trend of a stock. For example, there is an indicator, namely relative strength index (RSI). Calculated using average gain or loss of a stock during a time period, it suggests if the stock is overbought or oversold. Therefore, indicators like RSI can provide one more information about the current status of a stock than its volume traded and opening prices. C. Sentimental indicators These indicators reflect investors' optimism towards stock markets or the economy as a whole. They may also quantify the impact a particular social event or a change in the economy has on the markets. One example of these indicators is Consumer Confidence Index (CCI) conducted by The Conference Board in the United States. The index estimates households' emotions towards the current economy. Thus Indicators such as CCI can reflect willingness of people to invest in stock markets as they imply the wellness of the economy, which can affect the stock markets. These indicators will be filtered based on their impact on the accuracy of the classifier during experiments. Targeted companies: 12 European technology companies which are listed in NASDAQ. They are chosen because this amount of companies seems to be manageable and the stock of these companies are affected by similar news. This leads to another part of this project, which is to see how to improve the accuracy of the classifier with analysis on news and social media. Some sources of news and social media, such as Reuters and Twitter, may be used for the input to determine investors' sentiments towards the stocks. Therefore, they may enhance the prediction of the classifier. Judging from the complexity, the extent of enhancement made by them is going to be studied at a later stage of the project. The list of the 12 companies: 1. Amdocs Limited 2. ASML Holding N.V. 3. Atlassian Corporation Plc 4. Criteo S.A. 5. Ericsson 6. Materialise NV 7. Mimecast Limited 8. NXP Semiconductors N.V. 9. Seagate Technology PLC 10. Talend S.A. 11. trivago N.V. 12. Yandex N.V.