Dan Hislop

I recently faced a challenge to help a Customer Support department predict customer call volume: how many phone calls might arrive next month? The problem: customers have widely varying volumes, different seasonal patterns, and not always enough history.

This time-series analysis required seasonal decomposition in a way that addressed these challenges. By developing a portable metric called "Monthly Seasonal Percentage", I was able to move beyond individual customers to detect industry seasonality and with that, expand the projections to customers without year over year history. Read more about the project here.

• Blog Post

This notebook computes Mutual Information manually and compares outputs to sklearn's package. This confirms the math behind it and ensures understanding of the inner-workings of the function. This was an exercise under mentorship at work before using Mutual Info to evaluate features in a retention model.

Cases considered were: Continuous, Discrete (binary), and Discrete (not binary); Titanic test data was used.

• Notebook in Github

In this coding project I created a Random Variate Generator, which creates both discrete and continuous random variates from common distributions such as Geom(p), Exp(λ), Normal(μ,σ2), Gamma(α,β), Weibull(α,β), etc. Here I start with a uniform distribution and build the others, plotting my code's output against numpy's built-in generators.

The need for such a "classroom" project came into clear focus at work when I built a Bayesian Inference model to predict customer retention. Choosing the prior distributions were especially important given the small dataset I had (so it didn't overly influence outcomes), and this in-depth study paid off.

• Github

Predict player performance in Europe’s Big 5 Soccer Leagues, incorporating team play style and league effects. K-means clustering was used to group team play style, and Group Lasso Regression for predictions. (group project).

• Github
• Tableau Viz

Mapped Breweries per Capita using pandas, python, matplotlib, foursquare API and folium maps. A fun project = beer lover's West Coast travel guide.

• Jupyter Notebook
• Blog Post