Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis focuses on the contradiction between traditional finance and behaviour finance. Empirical irregularities observed in financial markets have caused behavioural finance to become very popular. This thesis zooms in on one of the most controversial anomalies in behavioural finance: overreaction on equity markets. Literature concerning both points of view is taken into account, but an especially deep overview is given of the evidence of overreaction by De Bondt and Thaler (1985, 1987).An empirical study is done to test overreaction on the Belgian equity market. To give our conclusions a strong foundation, additional overreaction tests were done on the NYSE, NASDAQ, OTC BB, NYSE Amex and Frankfurt exchanges.

Behavioural Finance. --- Belgium. --- Contrarian. --- De Bondt. --- Efficient market. --- Empirical study. --- Overreaction. --- Thaler. --- Traditional Finance. --- Underreaction.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Risk aversion has been treated by traditional finance as a stable input that was more useful to match the investors’ needs than a real driver for the financial markets. With the development of behavioural finance, some analysts tried to use this metric to improve their predictions over the markets. However, risk aversion is hard to aggregate. Most early attempts to find a relevant risk aversion of the market as a whole failed.
In the early 2000’s, researchers started focusing on the derivatives markets to improve the measurement of the representative investor’s attitude toward risk. Since hedges are possible over options, they can be used to create risk-neutral market expectations at a maturity. These expectations can be interpreted as probability density function by using different processes. Once these risk neutral density functions are created, it can be compared to the actual view of the representative investor over the market at the same maturity. The variation between the risk-neutral and the subjective densities can be interpreted as a proxy of the marginal rate of substitution at maturity. From this relation, the relative risk aversion of the market can be calculated. 
Unfortunately, the process to create the risk neutral density requires consequent amount of data. In this thesis, the focus is done on the creation of a coefficient that can be computed with less statistics. The information is not processed from every option price but from the implied volatility index of an underlying market index. Distribution is parametrized, and the risk aversion can be computed with only two variables per date. However, this simplification is lowering the accuracy of the created coefficient. In the second part of the thesis, the explanatory power of the simplified metric is evaluated through linear regressions. The goal is to assess the impact of the simplification process. 
The results are mixed. The created coefficients are able to explain a significant part of the returns of the index. But more could have been expected regarding the fact that this risk aversion metric is using ex-post information.

Risk aversion --- Implied volatility --- VIX --- risk-neutral density function --- Behavioural finance --- Sciences économiques & de gestion > Finance

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Over the past decade, computational methods, including machine learning (ML) and deep learning (DL), have been exponentially growing in their development of solutions in various domains, especially medicine, cybersecurity, finance, and education. While these applications of machine learning algorithms have been proven beneficial in various fields, many shortcomings have also been highlighted, such as the lack of benchmark datasets, the inability to learn from small datasets, the cost of architecture, adversarial attacks, and imbalanced datasets. On the other hand, new and emerging algorithms, such as deep learning, one-shot learning, continuous learning, and generative adversarial networks, have successfully solved various tasks in these fields. Therefore, applying these new methods to life-critical missions is crucial, as is measuring these less-traditional algorithms' success when used in these fields.

fintech --- financial technology --- blockchain --- deep learning --- regtech --- environment --- social sciences --- machine learning --- learning analytics --- student field forecasting --- imbalanced datasets --- explainable machine learning --- intelligent tutoring system --- adversarial machine learning --- transfer learning --- cognitive bias --- stock market --- behavioural finance --- investor’s profile --- Teheran Stock Exchange --- unsupervised learning --- clustering --- big data frameworks --- fault tolerance --- stream processing systems --- distributed frameworks --- Spark --- Hadoop --- Storm --- Samza --- Flink --- comparative analysis --- a survey --- data science --- educational data mining --- supervised learning --- secondary education --- academic performance --- text-to-SQL --- natural language processing --- database --- machine translation --- medical image segmentation --- convolutional neural networks --- SE block --- U-net --- DeepLabV3plus --- cyber-security --- medical services --- cyber-attacks --- data communication --- distributed ledger --- identity management --- RAFT --- HL7 --- electronic health record --- Hyperledger Composer --- cybersecurity --- password security --- browser security --- social media --- ANOVA --- SPSS --- internet of things --- cloud computing --- computational models --- metaheuristics --- phishing detection --- website phishing

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Over the past decade, computational methods, including machine learning (ML) and deep learning (DL), have been exponentially growing in their development of solutions in various domains, especially medicine, cybersecurity, finance, and education. While these applications of machine learning algorithms have been proven beneficial in various fields, many shortcomings have also been highlighted, such as the lack of benchmark datasets, the inability to learn from small datasets, the cost of architecture, adversarial attacks, and imbalanced datasets. On the other hand, new and emerging algorithms, such as deep learning, one-shot learning, continuous learning, and generative adversarial networks, have successfully solved various tasks in these fields. Therefore, applying these new methods to life-critical missions is crucial, as is measuring these less-traditional algorithms' success when used in these fields.

fintech --- financial technology --- blockchain --- deep learning --- regtech --- environment --- social sciences --- machine learning --- learning analytics --- student field forecasting --- imbalanced datasets --- explainable machine learning --- intelligent tutoring system --- adversarial machine learning --- transfer learning --- cognitive bias --- stock market --- behavioural finance --- investor’s profile --- Teheran Stock Exchange --- unsupervised learning --- clustering --- big data frameworks --- fault tolerance --- stream processing systems --- distributed frameworks --- Spark --- Hadoop --- Storm --- Samza --- Flink --- comparative analysis --- a survey --- data science --- educational data mining --- supervised learning --- secondary education --- academic performance --- text-to-SQL --- natural language processing --- database --- machine translation --- medical image segmentation --- convolutional neural networks --- SE block --- U-net --- DeepLabV3plus --- cyber-security --- medical services --- cyber-attacks --- data communication --- distributed ledger --- identity management --- RAFT --- HL7 --- electronic health record --- Hyperledger Composer --- cybersecurity --- password security --- browser security --- social media --- ANOVA --- SPSS --- internet of things --- cloud computing --- computational models --- metaheuristics --- phishing detection --- website phishing --- fintech --- financial technology --- blockchain --- deep learning --- regtech --- environment --- social sciences --- machine learning --- learning analytics --- student field forecasting --- imbalanced datasets --- explainable machine learning --- intelligent tutoring system --- adversarial machine learning --- transfer learning --- cognitive bias --- stock market --- behavioural finance --- investor’s profile --- Teheran Stock Exchange --- unsupervised learning --- clustering --- big data frameworks --- fault tolerance --- stream processing systems --- distributed frameworks --- Spark --- Hadoop --- Storm --- Samza --- Flink --- comparative analysis --- a survey --- data science --- educational data mining --- supervised learning --- secondary education --- academic performance --- text-to-SQL --- natural language processing --- database --- machine translation --- medical image segmentation --- convolutional neural networks --- SE block --- U-net --- DeepLabV3plus --- cyber-security --- medical services --- cyber-attacks --- data communication --- distributed ledger --- identity management --- RAFT --- HL7 --- electronic health record --- Hyperledger Composer --- cybersecurity --- password security --- browser security --- social media --- ANOVA --- SPSS --- internet of things --- cloud computing --- computational models --- metaheuristics --- phishing detection --- website phishing