Some of the material in is restricted to members of the community. By logging in, you may be able to gain additional access to certain collections or items. If you have questions about access or logging in, please use the form on the Contact Page.
In this dissertation, I explore different types of applications in the area of applied machine learning, time series analysis, and prediction. Time series forecasting is a fundamental task in machine learning and data mining. It is an active area of research, especially in applications that have direct impact on the real world. Foot traffic forecasting is one such application, which has a direct impact on businesses and non-profits alike. An accurate foot traffic prediction system can help retail businesses, physical stores, and restaurants optimize their labor schedule and costs, and reduce food wastage. In this work, we design a large scale data collection and prediction system for store foot traffic. We propose and compare different prediction models for foot traffic forecasting. Our foot traffic data has been collected from wireless access points deployed at over 65 businesses across the United States, for more than one year. We validate our work by comparing to state-of-the-art time series forecasting approaches. Results show the competitiveness of our proposed method in comparison to our previous work and state-of-the-art procedures for time series forecasting. Another challenging task in the area of time series forecasting is financial time series forecasting. As another part of my dissertation, I present a deep learning system for stock price prediction, which uses a variety of data for a subset of the stocks on the NASDAQ exchange to forecast the stock price. The prediction model is trained on the minutely data for a specific stock ticker and predicts the closing price of that stock ticker for multi-step-ahead. Our deep learning framework consists of a Variational Autoencoder for removing noise and uses time-series data engineering to combine the higher-level features with the original features. This new set of features is fed to a Stacked LSTM Autoencoder for multi-step-ahead prediction of the stock closing price. Besides, this prediction is used by a profit-maximization strategy to provide advice on the appropriate time for buying and selling a specific stock. Results show that the proposed framework outperforms the state-of-the-art time series forecasting approaches with respect to predictive accuracy and profitability. In the second part of my work, we present a web-based tool for automatic recoloring of web pages. Automatic application of different color palettes to web pages is essential for both professional and amateur web designers. However, existing recoloring tools for images and web pages do not provide full recoloring. We replace colors in .css, .html, and .svg files, and recolor images such as logos, banners, and background tiles to recolor web pages entirely. The new color theme is based on a color guide image provided by the user. The evaluation shows a high level of satisfaction with the quality of palettes and results of recoloring.