In this study, we conduct a detailed evaluation of machine learning and multimodal learning approaches in two distinct areas: a standard medical imaging benchmark and a novel material sciences benchmark. We utilize the CheXpert chest x-ray dataset for medical imaging and introduce a newly created Fluoropolymer Atomic Force Microscopy (AFM) dataset for material sciences. Both datasets are enhanced with additional images and binary metadata, encoded as one-hot vectors. We tested both pretrained and non-pretrained Convolutional Neural Network (CNN) models, such as ResNet50, ResNet101, DenseNet121, InceptionV3, and Xception, across different combinations of image and metadata inputs. Our results reveal that integrating multimodal data, including simple binary metadata, significantly enhances classification accuracy compared to conventional unimodal approaches or advanced MADDi models. This indicates the efficacy of multimodal learning in enriching data representation and boosting image classification performance. Notably, Xception models showed exceptional performance in CheXpert tests, and most models improved crystal structure predictions in AFM datasets. These insights set a new benchmark for performance and underscore the potential of multimodal learning in data-intensive applied science research.

US financial markets are influenced by complex interactions of diverse entities like institutional investors, which control a considerable portion of all US financial assets. Despite significant increase in the institutional ownership over the last several years, detection of causal associations between institutional investments and equity markets remains elusive due to inherent intricacies of the investment behavior. In this paper, we propose a novel solution to establish linkages between the institutional investments and market dynamics. We accomplish this task by deploying a multi-stage methodology that includes evaluation of heterogeneous data from disparate sources, an integrated framework comprising of tools to facilitate unstructured- and structured- data modeling, data integration, unsupervised learning, and an evaluation approach to uncover discernible patterns capturing market fragility. Our results on the real data confirm the efficacy of the proposed solution by establishing linkages between the investment behavior and market movements. For instance, the results show that the co-ownership and selling of large capitalization stocks held by institutional investors drive market returns co-movements. Our results confirm the efficacy of the presented framework. The proposed solution can assist the economists and policy makers detect fraud and proactively prepare against disruptive market movements thereby minimizing the risk to the economy.

Using big data, distributed computing systems such as Apache Hadoop requires processing massive amount of data to support business and research applications. Thus, it is critical to ensure the cyber security of such systems. To better defend from advanced cyber attacks that pose threats to even well-protected enterprises, system-auditing based techniques have been adopted for monitoring system activities and assisting attack investigation. In this demo, we are building a system that collects system auditing logs from a big data system and performs data analysis to understand how system auditing can be used more effectively to assist attack investigation on big systems. We also built a demo application that detects unexpected file deletion and presents root causes for the deletion.

US economy is driven by complex dynamics and interplay, often stemming from consumer spending, supply- demand associations, private market interactions, and government interventions. Intricacies within such a mixed economic ecosystem make it difficult to f orecast e vents l ike economic downturns. Recent financial c rises h ave s purred a d ebate over the efficacy o f e arly w arning s ystems, b ased o n traditional econometric modeling approaches, to forewarn against such events. Despite several advancements, accurate forecasting of these kinds of crises events remains elusive due to inherently complex interactions among the different economic constituents.In this paper, we propose a novel framework to predict economic crises events. We accomplish this task by deploying a multi-stage methodology that includes evaluation of macroeconomic indicators, a predictive modeling framework comprising of a diverse set of traditional and advanced forecasting models, and an evaluation approach to capture the contributions of different macroeconomic factors on multi-horizon forecasts, with good confidence. I n t his r esearch, w e l everage t he framework to forecast US economic downturns. Our results on the real test data outperform those from the survey of professional forecasters and confirm t he e fficacy of th e pr oposed ap proach to predict economic downturns ahead of multiple leading periods. The proposed solution can aid economists and policymakers prevent or proactively prepare against such events through appropriate monetary policies thereby minimizing the risk to the economy.

Financial markets are driven by complex dynamics and interplay, often stemming from convoluted investor interactions, asset and inter-market complexities. Such intricacies make it difficult to identify illegal trading activities like insider trading. Despite several advancements, the detection of such financial markets events remains elusive due to complex interactions among the market constituents. In this paper, we propose a novel solution to detect illegal trading activities driven by material nonpublic information. We accomplish this task by deploying a multistage methodology that includes a predictive modeling approach without the added constraint of having training data with the events of interest, an event prediction and detection methodology based on unstructured and structured data, a classification, and an evaluation approach to identify illegal insider trading events with good confidence. O ur r esults o n t he r eal t est d ata confirm the efficacy o f t he p roposed solution to detect insider trading activities in the U.S. equity markets.

The most important financial attribute of stock time series is the directionality of its price movement. This study found that the prediction error of neural network can only reflect the closeness between model predicted price and actual market price, but not the important financial attribute, the direction of stock price's ups and downs. Taking Chinese stock 600275 and American stock AMZN as examples, the former's prediction error is 0.0353, and its stock return rate is negative 59.49%, while the latter's prediction error is 0.0201, and its stock return rate is positive 26.49%, while the difference between the predictions of the two stocks is only 0.0152. It shows that under the same scale, absolute value constraint is the source of the problem that the prediction error has no stock up and down attributes. Therefore, in the absence of condition variable of trend direction state, it is unreliable to make the conclusion by judging the performance of regression neural network for stock price prediction based only on the value size of prediction error.