There are plenty of articles out there that explain what a decision tree is and what it does:

So here I am going to focus on how a decision tree may be implemented using the scikit-learn library in python on the iris dataset, along with some of the functionality that is useful in analysing the performance of the algorithm.

**What is a classifier?**

A classifier algorithm is used to map input data to a target variable through decision rules and can be used to predict and understand what characteristics are associated with a specific class or target. This means that…

Admittedly I am a huge fan of the NBA even though I am based in the UK so I don’t get to see much of the games. This means that I get my fix mostly from following the stats and the highlights after the games. Although I regret that I don’t get to watch as many games as I like, the analytical side of me enjoys being able to watch and follow the stats, usually being able to roll them off my tongue to any unsuspecting victim that engages me on the topic. Given this though, I thought it would…

Atul Gawande is an American surgeon, notable for his four best-selling books of ‘Complications’, ‘Better’, ‘Being Mortal: Medicine and What Matters in the End’ and ‘The Checklist Manifesto, a MacArthur Fellowship grant, and many other accolades that if you want to know more you can read his bio from the New Yorker.

While he is indeed a surgeon, and three of his bestselling books are about medicine, in 2009 he wrote his book ‘The Checklist Manifesto’ in which he describes the usefulness of checklists and why are beneficial in a variety of complicated and complex situations. The examples he provides…

Neural networks have been around for a long time, being developed in the 1960s as a way to simulate neural activity for the development of artificial intelligence systems. However, since then they have developed into a useful analytical tool often used in replace of, or in conjunction with, standard statistical models such as regression or classification as they can be used to predict or more a specific output. …

In February of this year, Tesla announced that it had purchased a large amount of Bitcoins (on the order of $1.5bn) and would now allow new Tesla’s to be bought with the Cryptocurrency¹. This came at a time when confidence in Crypto has been continually increasing but it seemed like this was the boost it needed to really take off. Along with Bitcoin itself, this move appeared to grant legitimacy to many Cryptocurrencies that had started to gain ground in potential mainstream circles, but a big move by Tesla has yet to have been made by any other company to…

In our attempt to cluster crimes in London in the previous article, we ignored the spatial dimension of the data in performing the clustering. Thus, this article seeks to remedy this by explicitly accounting for this.

Since the objective of the clustering was to identify how different clusters manifested themselves spatially, and that the original theory was that an LSOA located next to another is likely to be related in terms of the dominant type of crime, we need to account for this potential spatial relationship. This is done by imposing spatial constraints on clusters.

In doing so, this creates…

Following on from the previous article where the purpose of hierarchical clustering was introduced along with a broad description of how it works, the purpose of this article is to build on this by showing the results of its implementation through the sklearn package. As part of this I thought it would be good to combine this with a question of seeing how different types of crimes group together in London.

These crime clusters, identified by hierarchical clustering, are not the amount of crime but rather the type of crime in London. For example, are there areas in London where…

Clustering is a a part of machine learning called unsupervised learning. This means, that in contrast to supervised learning, we don’t have a specific target to aim for as our outcome variable is not predefined. For example, in regression or classification we already know what our model is trying to achieve by predicting a continuous target variable (y) or already defined classes. In contrast, for clustering we do not have a defined target variable, we are instead attempting to make those final groups ourselves, which we can then interpret in relation to our pre-existing knowledge.

This is useful in situations…

In my previous article I introduced the basic concepts of probability and how this relates to maximum likelihood estimation and ordinary least squares regression. In this article, I will continue to build on this foundation but in the case of Poisson regression through its application to the gravity model.

In contrast to linear regression, which assumes that the dependent variable is normally distributed with constant variance, Poisson regression assumes that the target variable distribution has a Poisson distribution. Thus, the variance of the independent variable is expected to equal to its mean. This is used in the case where we…

In my research I have come across the idea of maximum likelihood estimation quite a few times. However, without the statistical background of those that traditionally work in my field I often found it difficult to understand or follow some of the models they create and develop. A big part of this was their use of maximum likelihood estimation methods an their link to regression frameworks. …

A PhD student at the Centre for Advanced Spatial Analysis, UCL, looking to develop his coding skills! www.linkedin.com/in/philip-wilkinson-02917b150