Hello, hello!

Photo by Ed Robertson on Unsplash

Hello, and welcome!

Here’s a list of my articles, grouped into broad topics. So grab your favourite drink, get comfortable and dive in.

Helpful hints and tips

Articles about useful tips and tricks, from data analysis to making presentation-ready charts and exhibits.

• Create presentation-ready DataFrames: Comprehensive guide to formatting pandas DataFrames | Towards Data Science
• How to create stunning visualisations: Create presentation-ready line and bar charts in matplotlib | Towards Data Science
• Essential tools for your data manipulation tool box: Turbocharge your data manipulation skills | Towards Data Science
• Learn how to exploit Pandas’ agg function:Speed up data analysis using pandas agg | Towards Data Science
• Think Excel is the only place to find Pivot Tables? Think again: A how-to guide to pivot tables in pandas | Towards Data Science

Data science

Articles demonstrating various data science concepts, with real world data and applications.

• Reinsurance Pricing, the Data Science Way | by Bradley Stephen Shaw | Towards Data Science is a high-level explanation of alternative approachs to reinsurance pricing.
• Gaps in your knowledge of imputation techniques? Brdige them with Filling in the Gaps: Imputation 3 Ways | by Bradley Stephen Shaw | Towards Data Science.
• Seeing Numbers: Bayesian Optimisation of a LightGBM Model | by Bradley Stephen Shaw | Towards Data Science explores the tuning of a LightGBM model using Bayesian optimisation.
• Predicting Pulsar Stars: An Imbalanced Classification Task Comparing Bootstrap Resampling to SMOTE | by Bradley Stephen Shaw | Towards Data Science covers a typical data science workflow, summarising and comparing two popular approaches to dealing with class imbalance.
• The ins-and-outs of feature engineering for tabular data: Let’s Do: Feature Engineering. A brief exhibition of the power of… | by Bradley Stephen Shaw | Towards Data Science
• Let’s Do: Spatial Clustering with DBSCAN | by Bradley Stephen Shaw | Towards Data Science. A demonstration of a clustering tas using real-world data. Includes Bayesian optimisation of the DBSCAN clustering algorithm.
• Things all getting a little too serious for your liking? Take a break as we ask the question: Can data science find Bigfoot? | Towards Data Science

Neural networks

A documented journey of a dummy learning about neural networks, culminating in the building of a custom neural network to predict insurance claim frequency.

• Let’s Learn: Neural Nets #1. A step-by-step chronicle of me learning… | by Bradley Stephen Shaw | Medium
• Let’s Learn: Neural Nets #2 — Nodes and Neurons | Towards AI
• Let’s Learn: Neural Nets #3 — Activation Functions | Towards AI
• Let’s Learn: Neural Nets #4 — Weights and Biases | Python in Plain English
• Let’s Learn: Neural Nets #5 — Layers | by Bradley Stephen Shaw | Medium
• Let’s Learn: Neural Nets #6 — Backpropagation (or how Neural Networks learn) | by Bradley Stephen Shaw | Medium
• A neural network to predict insurance claim frequency | Towards Data Science: putting lessons 1–7 to the test, as I build a neural network to predict how often a customer will make an insurance claim.

Time series analysis and forecasting

Articles relating to all things time series.

• A comprehensive guide to time series decomposition | Towards AI breaks down how a time series can be… well, broken down.
• Familiar with “normal” cross-validation? Not quite sure if it’s appropriate for a time series problem? Let’s Do: Time Series Cross-Validation | Medium summarises why it isn’t and what you should do instead.
• False Prophet: Feature Engineering for Time Series | Towards Data Science. Taking inspiration from Meta’s Prophet, I explore how feature engineering can help us build a powerful time series model.
• False Prophet: a Time Series Regression Model | Towards Data Science combines inventive time series features with a practical model choice to build a powerful time series model. Inspired by Meta’s Prophet package, we explore how sometimes simple is better.
• Built a model following Prophet’s approach? Interested in how it stacks up to the real deal? I was — Comparing a Regression Model to Meta’s Prophet | Towards Data Science (medium.com)
• Sometimes it takes more than one tool to get the job done — see how that plays out in a time series context with Time Series: Mixed Model Time Series Regression | Towards Data Science
• Interested in how you might incorporate a new data source into a time series regression model? I go into some detail in External Data in Time Series Regression | Towards Data Science, using real-world weather in the prediction of road traffic incidents.
• How do you know that your time series model captures all of the signal in your target? Take a look at A comprehensive guide to time series residuals | Python in Plain English (medium.com), where I cover how looking at leftovers can help you build better models.

Happy reading!