|
|
Expert System for Decision-Making on Stock Markets Using Investor Sentiment
Janková, Zuzana ; Lenort, Radim (referee) ; Zinecker, Marek (referee) ; Chramcov, Bronislav (referee) ; Dostál, Petr (advisor)
The presented dissertation examines the potential of using the sentiment score extracted from textual data with historical stock index data to improve the performance of stock market prediction through the created model of the expert system. Given the large number of financial-related text documents published by both professional and amateur investors, not only on online social networks that could have an impact on real stock markets, but it is also crucial to analyze and in particular extract financial texts published by different users. investor sentiment. In this work, investor sentiment is obtained from online financial reports and contributions published on the financial social platform StockTwits. Sentiment scores are determined using a hybrid approach combining machine learning models with the teacher and neural networks, with multiple lexicons of positive and negative words used to classify sentiment polarity. The influence of sentiment score on the stock market through causality, cointegration and coherence is analyzed. The dissertation proposes a model of an expert system based on fuzzy logic methods. Fuzzy logic provides remarkable features when working with vague, inaccurate or unclear data and is able to deal with the chaotic environment of stock markets. In recent scientific studies, it has gained in popularity a higher level of fuzzy logic, which is referred to as type-2 fuzzy logic. Unlike the classic type-1 fuzzy logic, this higher type is able to integrate a certain level of uncertainty between the dual membership functions. However, this type of expert system is considerably neglected in the subject issue of stock market prediction using the extracted investor sentiment. For this reason, the dissertation examines the potential to use and the performance of type-2 fuzzy logic. Specifically, several type-2 fuzzy models are created. which are trained on historical stock index data and sentiment scores extracted from text data for the period 2018-2020. The created models are assessed to measure the prediction performance without sentiment and with the integration of investor sentiment. Subsequently, based on the created expert model, the investment strategy is determined, and its profitability is monitored. The prediction performance of fuzzy models is compared with the performance of several comparison models, including SVM, KNN, naive Bayes and others. It has been observed from experiments that fuzzy logic models are able to improve prediction by appropriate setting of membership and uncertainty functions contained in them and are able to compete with classical expert prediction models, which are standardly used in research studies. The created model should serve as a tool to support investment decisions for individual investors.
|
|
|
Machine learning in customer churn prediction
Fridrich, Martin ; Chramcov, Bronislav (referee) ; Lenort, Radim (referee) ; Šimberová, Iveta (referee) ; Dostál, Petr (advisor)
The dissertation examines customer churn prediction in e-commerce retail settings, presenting the current research landscape, analyzing key trends, and pinpointing opportunities for further investigation. The literature review is conducted using language processing. The study aims to develop, implement, and evaluate a machine learning system for predicting customer churn in the e-commerce environment, considering the economic implications of retention efforts, and facilitating a deeper understanding of the modeled phenomenon. The solution is organized into sections covering problem definition, data comprehension and processing, model development, evaluation, interpretation, and deployment. The author extends the traditional concept of customer churn as the lack of a transaction in a future period with a novel idea of the incremental economic impact of a retention campaign. The notions are validated using two datasets. The modeling framework incorporates GLM, SVM, ANN, decision trees, and meta-algorithms. Bayesian optimization estimates external parameters related to data processing and model building. The understanding of the phenomena is enhanced using SHAP tools, which are improved in terms of computation and visual representation. From the perspective of natural prediction performance, random forests and gradient boosting dominate; in the original task, ANN also performs well. When considering the financial results of the retention campaign, the novel approach functions excellently, mainly when coupled with decision trees or meta-learning. Recency and frequency representations of interactions and transactions are identified as key features; the feature importance of customer value emerges in the novel approach. Identifying and comprehending customer segments to target directly supports subsequent retention initiatives. In summary, the thesis offers an extensive overview of novel methods and tools for predicting customer churn, which can be valuable for future research and practical applications in business or educational settings.
|
|
|
The use of convolutional neural networks for predicting the financial failure of a company
Šebestová, Monika ; Chramcov, Bronislav (referee) ; Lenort, Radim (referee) ; Režňáková, Mária (referee) ; Dostál, Petr (advisor)
The doctoral thesis deals with the use of convolutional neural networks for predicting the financial failure of companies. A bibliometric analysis was used during the processing of the literature review, which enabled a better orientation in scientific works oriented to the methods and approaches used in the past to predict the financial failure of companies. On the basis of the obtained knowledge, a deep learning model based on the GoogLeNet architecture was proposed, with inputs consisting of financial and macroeconomic indicators of companies. The modeling was based on the transfer learning method, in which it is possible to fine-tune the parameters of the pre-established networks to accelerate the learning process of the convolutional neural network. The initial set of financial and macroeconomic indicators was compiled from the variables that were most often used in business failure prediction models in scientific papers. Appropriate statistical methods were used for the specific selection of indicators from which the model is built. Since convolutional neural networks work best with image processing, the quantitative values of the input indicators were graphically interpreted and it was investigated which type of graphical processing is most suitable for predicting the failure of companies. Due to the existence of an unbalanced data set, the effect of the SMOTE method on the accuracy of the model's prediction was analyzed in the thesis. The method was used to increase the number of samples of the minority class of firms. To model the prediction of financial default, several variants of models were proposed, which differed in the form of input data. It was tested how the removal of outliers from the data set, the point in time from which the data come or the method of predictor selection will affect the accuracy of the prediction. The parameters of the resulting model were further fine-tuned so that it was able to classify businesses from new real data. The research conducted showed that using the right type of graphical processing of input data, SMOTE technique and appropriate parameter settings, convolutional neural networks can predict the financial failure of companies with high accuracy.
|
|
|
Expert System for Decision-Making on Stock Markets Using Investor Sentiment
Janková, Zuzana ; Lenort, Radim (referee) ; Zinecker, Marek (referee) ; Chramcov, Bronislav (referee) ; Dostál, Petr (advisor)
The presented dissertation examines the potential of using the sentiment score extracted from textual data with historical stock index data to improve the performance of stock market prediction through the created model of the expert system. Given the large number of financial-related text documents published by both professional and amateur investors, not only on online social networks that could have an impact on real stock markets, but it is also crucial to analyze and in particular extract financial texts published by different users. investor sentiment. In this work, investor sentiment is obtained from online financial reports and contributions published on the financial social platform StockTwits. Sentiment scores are determined using a hybrid approach combining machine learning models with the teacher and neural networks, with multiple lexicons of positive and negative words used to classify sentiment polarity. The influence of sentiment score on the stock market through causality, cointegration and coherence is analyzed. The dissertation proposes a model of an expert system based on fuzzy logic methods. Fuzzy logic provides remarkable features when working with vague, inaccurate or unclear data and is able to deal with the chaotic environment of stock markets. In recent scientific studies, it has gained in popularity a higher level of fuzzy logic, which is referred to as type-2 fuzzy logic. Unlike the classic type-1 fuzzy logic, this higher type is able to integrate a certain level of uncertainty between the dual membership functions. However, this type of expert system is considerably neglected in the subject issue of stock market prediction using the extracted investor sentiment. For this reason, the dissertation examines the potential to use and the performance of type-2 fuzzy logic. Specifically, several type-2 fuzzy models are created. which are trained on historical stock index data and sentiment scores extracted from text data for the period 2018-2020. The created models are assessed to measure the prediction performance without sentiment and with the integration of investor sentiment. Subsequently, based on the created expert model, the investment strategy is determined, and its profitability is monitored. The prediction performance of fuzzy models is compared with the performance of several comparison models, including SVM, KNN, naive Bayes and others. It has been observed from experiments that fuzzy logic models are able to improve prediction by appropriate setting of membership and uncertainty functions contained in them and are able to compete with classical expert prediction models, which are standardly used in research studies. The created model should serve as a tool to support investment decisions for individual investors.
|
guest ::
