Grant: Discovery Grants Program - Individual

Detecting outliers (objects considerably dissimilar and inconsistent with respect to the majority of data) is an important and challenging research problem with a wide range of real-world applications in engineering, business, security, and healthcare. Intrusion detection in the Internet traffic, for instance, is a permanent concern of cybersecurity for Internet Service Providers. Similarly, detecting financial frauds has been the center of attention for regulators and investors since such frauds cause substantial damages to businesses and society. Detecting outliers, especially in large-scale unstructured data that includes both textual and numerical records, is a computationally complex problem that requires innovative solutions beyond traditional data analytics techniques. In this research program, I will explore, design, analyze and implement reliable AI algorithms for outlier detection in large-scale unstructured data using emerging word embedding techniques for converting texts to vectors of features, as well as dimension reduction and data summarization methods, and advanced AI and machine learning models (both text mining and data mining models) together with an ensemble of models, to make fast and accurate detection of outliers in vast amounts of documents containing both textual and numerical records. I will apply the new methods to the emerging applications in engineering (namely, detection of intrusion detection in computer networks), and business (namely, fraud detection in financial big data), protecting public and private organizations from a substantial socio-economic damages. Designing AI solutions that are unbiased, reliable, and trustworthy is another important research challenge that will be investigated in this research program at two levels: (a) data collection and pre-processing: the collected data must be inclusive, fair and trustworthy in order to be reliable. The collected and pre-processed data will be verified not to be biased towards specific organizations, regions, or minority groups; and (b) AI Algorithm design: the AI algorithms should not target specific organizations, regions, or minority groups either. In order to make reliable and trustworthy conclusions, the AI algorithms must remain fair, unbiased, and as much as possible, explainable.