About More Predictive Modeling Projects

You can now work through predictive modeling machine learning projects using Python. Now what? In this chapter, we look at ways that you can practice and refine your new found skills.

 1.1 Build And Maintain Recipes

 Throughout this book you have worked through many machine learning lessons using Python. Taken together, this is the start of your own private code base that you can use to jump-start your current or next machine learning project. These recipes are a beginning, not an end. The larger and more sophisticated that your catalog of machine learning recipes becomes, the faster you can get started on new projects and the more accurate the models that you can develop. As you apply your machine learning skills using the Python platform, you will develop experience and skills with new and different techniques with Python. You can pull out or abstract snippets and recipes as you go along and add them to your own collection of recipes, building upon the code that you can use on future machine learning projects. With time, you will amass your own mature and highly-tailored catalog of machine learning code for Python.

1.2 Small Projects on Small Datasets

Keep practicing your skills using Python. Datasets from the UCI Machine Learning Repository were used throughout this book to demonstrate how to achieve specific tasks in a machine learning project. They were also used in the longer case study projects. They are standardized, relatively clean, well understood and excellent for you to use as practice datasets.

You can use the datasets on the UCI Machine Learning repository as the focus of small (5-to-10 hours of effort) focused machine learning projects using the Python platform. Once completed, you can write-up your findings and share them online as part of your expanding portfolio of machine learning projects. These can be used by you later as a repository of knowledge on which you can build and further develop your skills. They can also be used to demonstrate to managers or future employers that you are capable of delivering results on predictive modeling machine learning projects using the Python platform. Here is a process that you can use to practice machine learning on Python:

• Browse the list of free datasets on the repository and download some that look interesting to you.
• Use the project template and recipes in this book to work through the dataset and develop an accurate model.
• Write up your work-flow and findings in a way that you can refer to them later or perhaps share it publicly on a website.

1.3 Competitive Machine Learning

Use competitive machine learning to push your skills. Working on small projects in the previous section is a good way to practice the fundamentals. At some point the problems will become easy for you. You also need to be pushed out of your comfort zone to help you grow your skills further. An excellent way to develop your machine learning skills with Python further is to start participating in competitions. In a competition, the organizer provides you with a training dataset, a test dataset on which you are to make predictions, a performance measure and a time limit. You and your competitors then work to create the most accurate model possible. Winners often get prize money. These competitions often last weeks to months and can be a lot of fun. They also offer a great opportunity to test your skills with machine learning tools on datasets that often require a lot of cleaning and preparation. The premier website for machine learning competitions is Kaggle2 . Competitions are stratified into different classes such as research, recruitment and 101 for beginners. A good place to start would be the beginner competitions as they are often less challenging and have a lot of help in the form of tutorials to get you started.

1.4 Summary

1. To continue to build up and maintain your catalog of machine learning recipes starting with the catalog of recipes provided as a bonus with this book.
2. To continue to work on the standard machine learning datasets on the UCI Machine Learning Repository.
3. To start work through the larger datasets from competitive machine learning and even start participating in machine learning competitions.