Social network platforms like Reddit are increasingly impacting real-world economics. Meme stocks are a recent phenomena where price movements are driven by retail investors organizing themselves via social networks. To study the impact of social networks on meme stocks, the first step is to analyze these networks. Going forward, predicting meme stocks' returns would require to predict dynamic interactions first. This is different from conventional link prediction, frequently applied in e.g. recommendation systems. For this task, it is essential to predict more complex interaction dynamics, such as the exact timing. These are crucial for linking the network to meme stock price movements. Dynamic graph embedding (DGE) has recently emerged as a promising approach for modeling dynamic graph-structured data. However, current negative sampling strategies, an important component of DGE, are designed for conventional dynamic link prediction and do not capture the specific patterns present in meme stock-related social networks. This limits the training and evaluation of DGE models in such social networks. To overcome this drawback, we propose novel negative sampling strategies based on the analysis of real meme stock-related social networks and financial knowledge. Our experiments show that the proposed negative sampling strategies can better evaluate and train DGE models targeted at meme stock-related social networks compared to existing baselines.

Modern technological developments, such as smart chips, sensors, and wireless networks, have revolutionized data-collection processes. One type of data that can be highly beneficial is event data due to the general conceptualization of an event. Monitoring event data enables real-time monitoring since an observation becomes available as soon as the event happens. Most of the available literature on monitoring event data is focused on vector-based time between events (TBEs) data. Methods for this type of data incorporate monitoring delays either due to overseeing temporal dependencies between variables or the need to wait until a complete vector is observed. To tackle these issues, we propose a multivariate monitoring scheme of event data that signals in real time. Our contribution is twofold: Our proposed method can monitor high-dimensional event data and it is computationally quick and easy to implement, thereby, outperforming existing methods that are only feasible to implement up to ten dimensions.

The food industry has been facing extreme shortages of food as a result of higher consumption due to increasing population. One of the key reasons of food shortages is inadequate water supply to the crops and plants. Usually farmers setup a schedule to supply water without assessing the real-time condition of the plants, which leads to wasting water in substantial quantities. As a result, plants are sometimes under watered or over watered. In this paper, we propose an IoT based watering system for plants that uses a microcontroller, a soil moisture sensor, an environmental temperature sensor, and a humidity sensor to assess favourable conditions of a plant’s growth. We propose 4 different experimental setups including regular watering schedules setups and modified watering schedule setups for both indoor and outdoor settings. We observe that watering the plants by a modified schedule based on the plants condition, the growth increases by using minimum amount of water. We further apply regression analysis on different variables in our system to observe the factors that have a direct effect on the growth of the plant. Based on our data analysis, environmental temperature plays the most important part in the growth of a plant along with adequate water supply.

Solar energy is the fastest growing source of renewable energy worldwide, and is set to grow at an unprecedented pace for the coming years. Large scale solar energy projects now compete with conventional energy production and offer attractive returns to investors. Solar energy projects have a projected lifetime of over 25 years and while the returns are attractive, investors rarely oversee the risks that impact their Return on Investment (ROI) over the long term. In the wake of increasingly fiercer competition among PV module manufacturers, quality often takes a backseat. In this project, we propose a prognostic solution contrary to existing reactive approaches. We develop a data-driven decision support system (DSS) for technical derisking of utility scale solar energy projects. This system can provide project stakeholders insight into risks associated to different manufacturers. The system is based on data gathered by Sinovoltaics Group, a leading solar quality assurance company with 10+ years of experience. Using information extraction algorithms, useful data is extracted from a large number of quality assurance reports from Sinovoltaics Group, compiled in a database and analyzed for risk assessment leading to a DSS.