Combining deep learning neural networks and spatiotemporal models for prediction and inference of residential house prices for property assessment

Property valuation is a crucial economic service that is used by local governments for the distribution of property taxes in order to fund local services. The current and widely adopted cost approach to valuation is based on estimating the land value and the depreciated cost of the building.

Development of a model for computational sea ice monitoring - Year Two

The proposed research project focuses on the development of a novel model for the computation of sea ice parameters in near real- time relying on satellite data. The interdisciplinary team will investigate solutions for high performance computing to monitor sea ice and calculate ice parameters with the high spatial resolution. This project includes R&D activities in sea ice modeling, calculating parameters of ocean interaction with sea ice and designing algorithms for satellite data processing and analysis.

Uncertainty Quantification for Deep Neural Networks

Deep neural networks are effective at image classification and other types of predictive tasks, achieving higher accuracy than conventional machine learning methods. However, unlike these other methods, the predictions are less interpretable. While accuracy may be enough for applications where errors are not costly, for real world applications, we want to also know when the predictions are more likely to be correct. Estimating the likelihood that a prediction is correct is called confidence, or uncertainty.

Modelling the Dependence between Loss Frequency and Loss Rate

Lending to various companies and individuals is a core business of banks. This lending activity comes with credit risk, namely the risk that some borrowers default and fail to make required payments. Estimating credit risk accurately is important for banks’ risk management. In this project, we analyze and model the dependence between loss frequency and loss rate of defaulting customers. The reason for the dependence comes from the underlying economic cycle: in an economic downturn, losses occur both more frequently and more severely than in an economic boom.

Non-convex learning with stochastic algorithms

In recent years, deep learning has led to unprecedented advances in a wide range of applications including natural language processing, reinforcement learning, and speech recognition. Despite the abundance of empirical evidence highlighting the success of neural networks, the theoretical properties of deep learning remain poorly understood and have been a subject of active investigation. One foundational aspect of deep learning that has garnered great intrigue in recent years is the generalization behavior of neural networks, that is, the ability of a neural network to perform on unseen data.

Enabling Purchase of Residential Homes at Scale

Properly buys and sells homes directly from consumers. For our business to be successful, we must be able to predict two things when making a home purchasing decision: ? The price it would sell for on the open market ? How much time it will spend on the market to sell at that price These two variables are correlated: price can affect time-on-market, and time-on-market can affect price. There are many other factors at play as well. In the broader market, these complex real estate decisions are largely made using human judgement, based on experience and expertise.

Probabilistic Transitive Closure of Fuzzy Cognitive Maps: Algorithm Enhancement and Application to Work Integrated Learning

Shopify has a well-developed partnership with universities for a work-integrated Bachelor of Computer Science degree. It is in their best interest to see their student interns successfully transition to the work place. In the application part of this study, we will use a fuzzy cognitive map (that is, a special type of a graph) to represent expert knowledge on determinants of success and well-being of a student intern, Moreover, a relatively new mathematical model – transitive closure – will be applied to analyze this data and compute a set of guidelines for better learning outcomes.

Evaluate effect of chemical compounds on plants using statistics and machine learning

To understand the effects of various substances on plants in terms of yield and disease severity (phytopathology), we need to evaluate both statistical significance and biological relevance when conducting biological experiments. Biological relevance refers to the nature and size of biological changes or differences seen in studies that would be considered relevant, while Statistically significance is the likelihood that a relationship between two or more variables is caused by something other than chance.

Statistical and Physiological Beat Modelling of Seismocardiogram Signal - Year 2

"Seismocardiogram (SCG) is a signal that is captured by placing an accelerometer on the human chest. This signal captures very important timing information such as opening and closing of the heart valves. In addition to these timing information, the non-invasive nature of this signal makes it an attractive solution for remote monitoring of patients with heart conditions.
The morphology of SCG signal changes depending on different types of heart conditions and diseases. A mathematical model represents the morphology of a signal in terms of certain parameters.

A behavioural risk model for deposit only customers

In this joint collaboration with BNS, we will develop a behavioural risk model to predict the likelihood of future risk of breaking the promise to pay debt for customers who only hold deposit products with BNS. The model will be utilized to support business operations such as credit card and loan pre-approvals. That is to say, if you are a customer who only have chequing, saving and/or investment accounts with BNS and plan to buy a car, you will be scored in this model for the car loan pre-approval.