Here I am gathering my work on the Melbourne Pedestrian footfall time-series.
-
Data Exploration Initial exploration of the Melbourne dataset, locations of sensors, missing data.
-
Distributions and Hypothesis testing.
- 2a Distributions and Hypothesis testing What distributions should the pedestrian counts follow?
- 2b Poisson distributions, Negative Binomial and Normal Distribution Explanation of why we expect the Poisson distribution to tend to a normal distribution when the counts are large. Why might we actually be deviating from the Poisson distribution and seeing a negative binomial distribution.
-
PCA Analysis.
- 3a PCA Analysis Can apply PCA to reduce a days/weeks worth of data down to a smaller set of numbers? What can we learn about the usage of the space? It turns out we can learn quite a lot.
- 3b PCA Outline Short outline of what PCA does for those unfamiliar.
- 3c PCA Analysis - As a time-series. Looking at the first components of the PCA can we use this to visualise changes over time? Particularly after COVID?
-
Clustering It would be great to be able to cluster these locations together. Do we have true labels for the sites? Sort of, we have the location names and with a little work we can use these to grade the quality of our clustering analysis.
-
Forecasting
- 5a Forecasting Introduction Introduction to some of the key concepts of building forecasting models.
- 5b Forecasting Starting to build forecasting models and setting a baseline.