Michael Grogan

Michael Grogan is a data scientist with expertise in TensorFlow and time series analysis. His educational background is a master’s degree in economics from University College Cork, Ireland. As such, much of his work has been in the domain of business intelligence; i.e. using machine learning technologies to develop solutions to a wide range of business problems. He has implemented time series solutions for organizations across a range of industries through implementation of statistical analysis as well as more advanced machine learning methodologies. In addition, he has delivered numerous seminars and training courses in the areas of data science and machine learning, including for Manning and O'Reilly Media.

projects by Michael Grogan

Time Series Forecasting with Bayesian Modeling

5 weeks · 4-6 hours per week average · ADVANCED

Bayesian-based probability and time series methods allow data scientists to adapt their models to uncertainty and better predict outcomes. In this series of liveProjects, you’ll take on the role of a data scientist making customer predictions for hotels and airlines. You’ll use ARIMA, Bayesian dynamic linear modeling, PyMC3 and TensorFlow Probability to model hotel booking cancelations, and implement a Prophet model with uncertainty analysis to forecast air passenger numbers. Each project in the series is focused on a different time series forecasting model, allowing you to compare model performance and choose the skills most relevant to your career development.

Time Series Modeling with TensorFlow Probability

1 week · 4-6 hours per week · ADVANCED

In this liveProject, you’ll combine the power of deep learning with probabilistic modeling. You’ll build a structural time series model that can develop probabilistic forecasts of hotel cancellations, and use this model to identify anomalies across your cancellation data. You’ll perform a similar analysis of an air passenger dataset, and then use Bayesian Switchpoint analysis to determine the approximate time interval in which searches for the term “vacation” declined.

Prophet Model Incorporated with Bayesian Analysis

1 week · 4-6 hours per week · ADVANCED

In this liveProject, you’ll build a Prophet model that can forecast airline passenger numbers using data from the DataSF portal. The hotel you work for believes that analyzing the travel trends of US customers will help them forecast potential travel to Europe, and bookings in the hotel. You’ll enhance your model with "changepoints" that mark a significant change in trends, and make adjustments so your model can account for uncertainty in the trend and seasonality components.

Bayesian Statistical Methods with PyMC3

1 week · 4-6 hours per week · ADVANCED

In this liveProject, you’ll use PyMC3 to generate a posterior distribution of hotel cancellations. This will allow you to build a predictive model that can update its predicted probabilities by incorporating new information. You’ll master methods for modelling mean and standard deviations based on the selected prior values, generating distributions to determine if data follows an AR(1) process, modeling of stochastic volatility and generalized linear modelling with PyMC3.

Bayesian Dynamic Linear Modeling

1 week · 4-6 hours per week · ADVANCED

In this liveProject, you’ll build a Bayesian dynamic linear model that can take account of sudden state space changes and rapidly react to dramatic trend changes. These trend changes could take many forms—from heightened demand during a major sporting event, to a global pandemic that causes cancellations to skyrocket. You’ll use the PyDLM library to generate forecasts that can dynamically adapt to the unforeseen, and quickly shift to making accurate predictions for a new reality.

Data Manipulation and ARIMA Modeling with Pyramid

1 week · 4-6 hours per week · ADVANCED

In this liveProject, you’ll investigate seasonality in hotel cancellations by building an ARIMA model that can predict cancellations on a weekly basis. You’ll learn how to manipulate a dataset with pandas in order to form a weekly time series, before going on to make your first predictions.