Kaggle (acquired by Google in 20217) is a great resource for all data scientists. The company touts itself as the “home of data science.” UCLA’s introductory data science students are encouraged to first complete the venerable Titanic and take advantage of his Kaggle. Introduction Prediction Challenge, and move on to active challenges. Kaggle is a great way to gain valuable experience in data science and machine learning. Well, here are two of his excellent books that guide you through the Kaggle process. The Kaggle Book by Konrad Banachewicz and Luca Massaron, published in 2022, and The Kaggle Workbook by the same authors, published in 2023, are both published by UK-based Packt Publishing.
let’s get started Kaggle book. This book is an invaluable learning resource not only for those entering Kaggle competitions, but for almost any data scientist looking to hone their skills. Reading this book is like merging the minds of a Kaggle Master or Grandmaster and a Vulcan. You can immediately see how these professionals are doing well in his Kaggle ecosystem. This is accomplished in a number of ways, including award-winning Python code, in-depth interview sidebars spread throughout the book, and curated links pointing to key discussions on Kaggle. The last feature of the book is probably the most useful, as it also has discussions that provide insights you won’t find anywhere else. For example, Grand Master Michael Jahrer’s famous post on denoising autoencoders is presented in Chapter 7. Porto reading his detailed explanation of how he won first place in his Seguro safe driving prediction contest is an excellent way to add his own data to his science toolbox. is. Chapter 7 also includes an insightful interview with the famous XGBoost enthusiast Kaggler Bojan Tunguz. twitter.
The book also provides strategic references to many Kaggle competitions that describe important ways to ensure machine learning success. For example, Chapter 5 contains references to numerous contests where AUC was used to determine classification accuracy. Move happily from project to project to gain a deeper understanding of key machine learning principles. This book serves as a guide map for such quests. As a result, you will have a better understanding of how to approach future projects.
One of my favorite chapters is Chapter 5 on metrics. Ultimately, you need a solid technique for judging the performance of your ML solution. Another favorite is Chapter 8 on hyperparameter optimization. Having the best and most powerful algorithm is one thing, but knowing how to optimize the many hyperparameters of your model is quite another. This book does not cover the mathematical underpinnings of algorithms and their hyperparameters, but does provide some insight into finding the best hyperparameters for your model. It will be of great value to see how Grandmaster deals with the hyperparameter problem. I also enjoyed Chapter 7 on modeling tabular data, or business data. It covers important topics such as dimensionality reduction, feature engineering, and using neural networks for tabular data.
The rest of the book includes helpful topics such as an introduction to Kaggle datasets, working with Kaggle notebooks, working with Kaggle discussion forums, as well as popular topics such as computer vision and NLP. This book is a great way to control the complex Kaggle infrastructure, and I can’t imagine running a Kaggle contest without it.
A very nice accessory for the Kaggle Book is Kaggle workbook Consisting of just four chapters, each chapter includes an in-depth review of past Kaggle challenges and can be viewed as a self-study exercise with valuable insights for the Kaggle data science competition. Each of the four chapters contains the Python source code for the solution. This code is designed to run on a Kaggle notebook. Here is the list of projects:
- Porto Seguro Safe Driver Prediction – Predict whether drivers will file an insurance claim next year. This project includes using the Light GBM model, building a denoising autoencoder and using it to feed a neural network, and blending the model.
- Kaggle’s M5 for Accuracy and Uncertainty – Based on Walmart’s daily sales time series of merchandise hierarchically arranged into departments, categories, and stores across three US states, this solution It shows how to use LightGBM for sequential problems.
- Cassava Leaf Disease Classification – Classify crowdsourced cassava plant photos. This multiclass problem shows how to build a complete pipeline for image classification.
- Google Quest Q&A Labeling – Predicts human respondents’ ratings of subjective aspects of question-answer pairs where contextual understanding is important. Cast as a multiclass classification problem, this solution explores the semantic properties of corpora.
Conclusion
If you’re thinking of joining a Kaggle challenge, or just want to improve your data science skills, I highly recommend this Kaggle book tandem. It is unthinkable to invest in one book and not in another. Both are required. These are great one-two punches to gain valuable experience solving machine learning problems.
Contributed by Daniel D. Gutierrez, Editor-in-Chief and Resident Data Scientist at insideBIGDATA. In addition to being a technology journalist, Daniel is also a data scientist consultant, author, educator, and serves on numerous advisory boards for various start-ups.
Sign up for the free insideBIGDATA newsletter.
Join us on Twitter: https://twitter.com/InsideBigData1
Join us on LinkedIn: https://www.linkedin.com/company/insidebigdata/
Join us on Facebook: https://www.facebook.com/insideBIGDATANOW
