Detecting Company-Specific Purchase Evidence from Twitter Posts

Delphia’s business model revolves around generating insights for investing firms that allow them to make better trading decisions. It has been shown that detecting when Twitter users post about recent or future purchases has the potential to increase the accuracy of company sales forecasts, which in turn can inform stock trading strategies. This internship project aims to develop automated means to detect and quantify purchase related posts on Twitter.

Multi-morbidity Characterization and Polypharmacy Side Effect Detection for designing Optimal Personalized Healthcare with Machine Learning

Despite a significant improvement in healthcare systems over the past decades, the rapid growth in the number of patients with multiple chronic diseases – called multimorbidity – stands as a complex challenge to healthcare services that are primarily designed to treat individuals with single conditions. Advances in machine learning as well as in computing power now enable us to exploit a vast amount of healthcare data.

Applying Machine Learning to Predict Demand Transference

The project will help us design a machine learning model that can determine the demand transference of our customers. The key objective of this project is to design, research, build, and experiment with machine learning models to ensure low product waste and high customer satisfaction. The model will have several impactful applications across the organization.

Machine Learning for Cancer Treatment Plan Benchmarking

Oncology specialists are few in numbers and cannot be present within every hospital or clinic providing their guidance and support. There are institutions in this world with oncology experts that can provide the best possible care and the knowledge of these experts is stored in medical case files within the hospital. With the power of machine learning and the internet, an organization without any experts can compare their cancer treatment plans to the top performing institutions in the world and receive constructive feedback on where their plan needs improvement.

Multi-Channel User Linkage through Probabilistic Matching

People utilize multiple devices to complete various tasks, making their online identities fragmented. Advertising is as much about knowing when not to promote a product as it is about when to do it. For example, before being sent alcohol and cannabis ads, the user must be identifiable as being over 19. Age information may only be available on a channel different than the one through which the user is connecting. This makes it difficult to gain a holistic understanding of users and develop a single marketing strategy across devices.

Predicting Risk of Aggressive Responsive Behaviours among People Suffering from Dementia using Natural Language Processing (NLP) and Machine Learning (ML).

Patients with dementia will eventually experience significant loss of cognitive function. Many will have difficulty properly communicating life’s challenges and instead become agitated, resulting in verbal or physical aggression. Monitoring the risk of a resident harming themselves or others due to aggressive behavior is a priority within a long-term care facility where dementia is present. Caregivers at Shannex regularly record resident health and behaviour using computing systems.

Application of Real-time Axis-based SLAM in Geotechnical Engineering

Mining operations evaluate tunnel stability by looking at orientations of exposed rock faces and extrapolating the underlying rock formation. The most widespread and traditional method for determining these characteristics involve hand measurements with static tools. This process is both time intensive and prone to error. Additionally, it may be dangerous to measure rock faces that are hard to reach. The goal of this project is to implement a novel algorithm that can quickly and automatically determine sets of parallel rock faces and their orientations.

Fast Implementation of Machine Learning Algorithms for Event Sequence Data in order to improve Customer Experience

Every day millions of customers move through the sales cycles of companies, this generates large sources of event data. This project aims to discover, understand and predict the journeys of their customers. On one hand, the project is interested in describing the data at a higher level. This means to apply machine learning techniques, namely clustering, and sequence embedding, in order to group similar behaviors together and allow the user to focus analysis on different aspects of the data, such as users of a specific age.

Machine Learning Aided Detection of Brain Aneurysms

Intracranial aneurysms are relatively common, occurring in 2-5% of the general population. Rupture of the aneurysm can result in a stroke with a devastating 30-day mortality of 45%. Further, severe medical conditions are possible in which up to one third of patients may die before reaching the hospital and one third will become severely debilitated. Aneurysms are, however, difficult to find on magnetic resonance and computed tomography angiography scans (MRA and CTA scans) – especially when small or located close to the bone.

Automated Retail Area Cluster Detection

The main objective of this project is to develop a retail cluster detection algorithm and improve the accuracy of the identification. The final deliverable will be an algorithm that runs through Google dataflow that will be able to ingest a month of users’ location breadcrumbs and output user-location clusters. The output of the algorithm will be a unique cluster identifier that will be used in the visit algorithm for visit identification. Besides latitude and longitude data, the project will have access to altitude information.