Given the volatility of the stock market and the multitude of financial variables at play, forecasting the value of stocks can be a challenging task. Nonetheless, such prediction task presents a fascinating problem to solve using machine learning. The stock market can be affected by news events, social media posts, political changes, investor emotions, and the general economy among other factors. Predicting the stock value of a company by simply using financial stock data of its price may be insufficient to give an accurate prediction. Investors often openly express their attitudes towards various stocks on social medial platforms. Hence, combining sentiment analysis from social media and the financial stock value of a company may yield more accurate predictions. This thesis proposes a method to predict the stock market using sentiment analysis and financial stock data. To estimate the sentiment in social media posts, we use an ensemble-based model that leverages Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN) models. We use an LSTM model for the financial stock prediction. The models are trained on the AAPL, CSCO, IBM, and MSFT stocks, utilizing a combination of the financial stock data and sentiment extracted from social media posts on Twitter between the years 2015-2019. Our experimental results show that the combination of the financial and sentiment information can improve the stock market prediction performance. The proposed solution has achieved a prediction performance of 74.3%.

The thesis focuses on the use of deep recurrent neural network, architecture Long Short-Term Memory for robust denoising of audio signal. LSTM is currently very attractive due to its characteristics to remember previous weights, or edit them not only according to the used algorithms, but also by examining changes in neighboring cells. The work describes the selection of the initial dataset and used noise along with the creation of optimal test data. For network training, the KERAS framework for Python is selected. Candidate networks for possible solutions are explored and described, followed by several experiments to determine the true behavior of the neural network.

In industry, it is crucial to be able to detect damage or abnormal behavior in machines. A machine's downtime can be minimized by detecting and repairing faulty components of the machine as early as possible. It is, however, economically inefficient and labor-intensive to detect machine fault sounds manual. In comparison with manual machine failure detection, automatic failure detection systems can reduce operating and personnel costs.  Although prior research has identified many methods to detect failures in drill machines using vibration or sound signals, this field still remains many challenges. Most previous research using machine learning techniques has been based on features that are extracted manually from the raw sound signals and classified using conventional classifiers (SVM, Gaussian mixture model, etc.). However, manual extraction and selection of features may be tedious for researchers, and their choices may be biased because it is difficult to identify which features are good and contain an essential description of sounds for classification. Recent studies have used LSTM, end-to-end 1D CNN, and 2D CNN as classifiers for classification, but these have limited accuracy for machine failure detection. Besides, machine failure occurs very rarely in the data. Moreover, the sounds in the real-world dataset have complex waveforms and usually are a combination of noise and sound presented at the same time. Given that drill failure detection is essential to apply in the industry to detect failures in machines, I felt compelled to propose a system that can detect anomalies in the drill machine effectively, especially for a small dataset. This thesis proposed modern artificial intelligence methods for the detection of drill failures using drill sounds provided by Valmet AB. Instead of using raw sound signals, the image representations of sound signals (Mel spectrograms and log-Mel spectrograms) were used as the input of my proposed models. For feature extraction, I proposed using deep learning 2-D convolutional neural networks (2D-CNN) to extract features from image representations of sound signals. To classify three classes in the dataset from Valmet AB (anomalous sounds, normal sounds, and irrelevant sounds), I proposed either using conventional machine learning classifiers (KNN, SVM, and linear discriminant) or a recurrent neural network (long short-term memory). For using conventional machine learning methods as classifiers, pre-trained VGG19 was used to extract features and neighborhood component analysis (NCA) as the feature selection. For using long short-term memory (LSTM), a small 2D-CNN was proposed to extract features and used an attention layer after LSTM to focus on the anomaly of the sound when the drill changes from normal to the broken state. Thus, my findings will allow readers to detect anomalies in drill machines better and develop a more cost-effective system that can be conducted well on a small dataset. There is always background noise and acoustic noise in sounds, which affect the accuracy of the classification system. My hypothesis was that noise suppression methods would improve the sound classification application's accuracy. The result of my research is a sound separation method using short-time Fourier transform (STFT) frames with overlapped content. Unlike traditional STFT conversion, in which every sound is converted into one image, a different approach is taken. In contrast, splitting the signal into many STFT frames can improve the accuracy of model prediction by increasing the variability of the data. Images of these frames separated into clean and noisy ones are saved as images, and subsequently fed into a pre-trained CNN for classification. This enables the classifier to become robust to noise. The FSDNoisy18k dataset is chosen in order to demonstrate the efficiency of the proposed method. In experiments using the proposed approach, 94.14 percent of 21 classes were classified successfully, including 20 classes of sound events and a noisy class. / <p>Vid tidpunkten för disputationen var följande delarbeten opublicerade: delarbete 2 och 3 inskickat.</p><p>At the time of the doctoral defence the following papers were unpublished: paper 2 and 3 submitted.</p> / AISound – Akustisk sensoruppsättning för AI-övervakningssystem / MiLo — miljön i kontrolloopen

The thesis focuses on the use of deep recurrent neural network, architecture Long Short-Term Memory for robust denoising of audio signal. LSTM is currently very attractive due to its characteristics to remember previous weights, or edit them not only according to the used algorithms, but also by examining changes in neighboring cells. The work describes the selection of the initial dataset and used noise along with the creation of optimal test data. For creation of the training network is selected KERAS framework for Python and are explored and discussed possible candidates for viable solutions.

In this thesis, methods of implementing an online gesture recognition system for the novel Stylaero Board device are investigated. Two methods are evaluated - one based on LSTMs and one based on HMMs - on three kinds of gestures: Tap, circle, and flick motions. A method’s performance was measured in its accuracy in determining both whether any of the above listed gestures were performed and, if so, which gesture, in an online single-pass scenario. Insight was acquired regarding the technical challenges and possible solutions to the online aspect of the problem. Poor performance was, however, observed in both methods, with a likely culprit identified as low quality of training data, due to an arduous and complex gesture performance capturing process. Further research improving on the process of gathering data is suggested.

This study has investigated whether it is possible to classify time series data originating from a gamified learning management system. By using the school data provided by the gamification company Insert Coin AB, the aim was to distribute the teacher’s supervision more efficiently among students who are more likely to fail. Motivating this is the possibility that the student retention and completion rate can be increased. This was done by using Long short-term memory and convolutional neural networks and Markov chain to classify time series of event data. Since the classes are balanced the classification was evaluated using only the accuracy metric. The results for the neural networks show positive results but overfitting seems to occur strongly for the convolutional network and less so for the Long short-term memory network. The Markov chain show potential but further work is needed to mitigate the problem of a strong correlation between sequence length and likelihood.

No description available.

Photoplethysmography (PPG) is a method to detect blood volume changes in every heartbeat. The peaks in the PPG signal corresponds to the electrical impulses sent by the heart. The duration between each heartbeat varies, and these variances are better known as heart rate variability (HRV). Thus, finding peaks correctly from PPG signals provides the opportunity to measure an accurate HRV. Additional research indicates that deep learning approaches can extract HRV from a PPG signal with significantly greater accuracy compared to other traditional methods. In this study, deep learning classifiers were built to detect peaks in a noise-contaminated PPG signal and to recognize the performed activity during the data recording. The dataset used in this study is provided by the PhysioBank database consisting of synchronized PPG-, acceleration- and gyro data. The models investigated in this study were limited toa one-layer LSTM network with six varying numbers of neurons and four different window sizes. The most accurate model for the peak classification was the model consisting of 256 neurons and a window size of 15 time steps, with a Matthews correlation coefficient (MCC) of 0.74. The model consisted of64 neurons and a window duration of 1.25 seconds resulted in the most accurate activity classification, with an MCC score of 0.63. Concludingly, more optimization of a deep learning approach could lead to promising accuracy on peak detection and thus an accurate measurement of HRV. The probable cause for the low accuracy of the activity classification problem is the limited data used in this study.

With an increasing demand for rapid high-capacity internet, the telecommunication industry is constantly driven to explore and develop new technologies to ensure stable and reliable networks. To provide a competitive internet service in this growing market, proactive detection and prevention of disturbances are key elements for an operator. Therefore, analyzing network traffic for forecasting disturbances is a well-researched area. This study explores the advantages and drawbacks of implementing a long short-term memory model for forecasting codeword errors in a hybrid fiber-coaxial network. Also, the impact of using multivariate and univariate data for training the model is explored. The performance of the long short-term memory model is compared with a multilayer perceptron model. Analysis of the results shows that the long short-term model, in the vast majority of the tests, performs better than the multilayer perceptron model. This result aligns with the hypothesis, that the long short-term memory model’s ability to handle sequential data would be superior to the multilayer perceptron. However, the difference in performance between the models varies significantly based on the characteristics of the used data set. On the set with heavy fluctuations in the sequential data, the long short-term memory model performs on average 44% better. When training the models on data sets with longer sequences of similar values and with less volatile fluctuations, the results are much more alike. The long short-term model still achieves a lower error on most tests, but the difference is never larger than 7%. If a low error is the sole criterion, the long short-term model is the overall superior model. However, in a production environment, factors such as data storage capacity and model complexity should be taken into consideration. When training the models on multivariate and univariate datasets, the results are unambiguous. When training on all three features, ratios of uncorrectable and correctable codewords, and signal-to-noise ratio, the models always perform better. That is, compared to using uncorrectable codewords as the only training data. This aligns with the hypothesis, which is based on the know-how of hybrid fiber-coaxial experts, that correctable codewords and signal-to-noise ratio have an impact on the occurrence of uncorrectable codewords. / På grund av den ökade efterfrågan av högkvalitativt internet, så drivs telekomindustrin till att konsekvent utforska och utveckla nya teknologier som kan säkerställa stabila och pålitliga nätverk. För att kunna erbjuda konkurrenskraftiga internettjänster, måste operatörerna kunna förutse och förhindra störningar i nätverken. Därför är forskningen kring hur man analyserar och förutser störningar i ett nätverk ett väl exploaterat område. Denna studie undersökte för- och nackdelar med att använda en long short-term memory (LSTM) för att förutse kodordsfel i ett hybridfiber-koaxialt nätverk. Utöver detta undersöktes även hur multidimensionell träningsdata påverkade prestandan. I jämförelsesyfte användes en multilayer perceptron (MLP) och dess resultat. Analysen av resultaten visade att LSTM-modellen presterade bättre än MLP-modellen i majoriteten av de utförda testerna. Men skillnaden i prestanda varierade kraftigt, beroende på vilken datauppsättning som användes vid träning och testning av modellerna. Slutsatsen av detta är att i denna studie så är LSTM den bästa modellen, men att det inte går att säga att LSTM presterar bättre på en godtycklig datauppsättning. Båda modellerna presterade bättre när de tränades på multidimensionell data. Vidare forskning krävs för att kunna determinera om LSTM är den mest självklara modellen för att förutse kodordsfel i ett hybridfiber-koaxialt nätverk.

Artificial Intelligence (AI) and Computer Vision (CV) have brought the promise of many applications along with many challenges to solve. The majority of current AI research has been dedicated to single-modal data processing meaning they use only one modality such as visual recognition or text recognition. However, real-world challenges are often a combination of different modalities of data such as text, audio and images. This thesis focuses on solving the Visual Question Answering (VQA) problem which is a significant multi-modal challenge. VQA is defined as a computer vision system that when given a question about an image will answer based on an understanding of both the question and image. The goal is improving the training time of VQA models. In this thesis, Look, Read, Reason and Answer (LoRRA), which is a state-of-the-art architecture, is used as the base model. Then, Reduce Uni-modal Biases (RUBi) is applied to this model to reduce the importance of uni- modal biases in training. Finally, an early stopping strategy is employed to stop the training process once the model accuracy has converged to prevent the model from overfitting. Numerical results are presented which show that training LoRRA with RUBi and early stopping can converge in less than 5 hours. The impact of batch size, learning rate and warm up hyper parameters is also investigated and experimental results are presented. / Graduate