Inc / Hodson Applied Machine Learning

Practical Models for Solving Real-World Business Problems
1. Auflage 2026
ISBN: 978-1-80865-832-7
Verlag: Packt Publishing
Format: EPUB
Kopierschutz: 0 - No protection

E-Book, Englisch, 443 Seiten

Business-focused machine learning begins with a clear view of the data and the decision it must support. The opening material establishes practical tools such as GitHub and Anaconda, introduces three use cases with dedicated datasets, and shows how visualization, descriptive statistics, correlation analysis, cleaning, and dummy coding shape dependable inputs.
The discussion then moves through a structured model-selection process. Readers compare regression, decision trees, random forests, gradient boosting, and clustering, while learning when each approach fits a business need. Validation metrics, interpretability, and iterative feature engineering provide a disciplined way to judge results, expose weak assumptions, and refine performance without treating the model as a black box.
The final stage connects analysis to operations through implementation, monitoring, prediction workflows, and impact measurement. Readers see how model quality must be maintained after launch and how outcomes can be linked to business value. By the end of this journey, readers can prepare data, choose and evaluate suitable models, and manage machine learning solutions from initial idea through long-term use.

Autoren/Hrsg.

Inc, Rheinwerk Publishing

Hodson, Jason

Leseproben

1.2    Learning to Google (or Prompt)

An underlying skillset in leveraging machine learning is learning how to search for the right answer to your problem. Traditionally, this meant learning how to Google—knowing which websites were the best references and how to navigate their respective interfaces. With the emergence of LLMs as a search tool, you now have more options, as we’ll discuss in this section.

1.2.1    What Can You Find with Google?

There are a lot of resources online, so knowing which resources are probabilistically the best is important. The Google algorithm may do some of this for you, but knowing which websites to focus on will help expedite your research and troubleshooting process. For example, some resources help you gain a general understanding of the issue while others are better at providing answers for specific use cases or questions. The resources that we’ll discuss are by no means extensive, but you’ll likely see them come up on a recurring basis.

We’ll use a simple search of “how to build random forest model sklearn” as an example. When you search for this on Google, you get the results shown in Figure 1.1. We’ll break down these results in the following sections.

Figure 1.1     Google Search for Random Forest Model

Scikit-learn: Python’s Machine Learning Swiss Army Knife

If you aren’t familiar with scikit-learn (also known as sklearn), that will change. sklearn is the most common machine learning library in Python. It has a vast array of tools and models, making it a great tool for learning the different applications of machine learning. The models we’ll explore in this book are all from sklearn, which has some significant benefits that you’ll see later in the book.

Generative AI Result

Even as of writing this book in 2025, the generative AI results are impressive. They give you all the necessary code to build a random forest model. However, a word of caution here: Generative AI is not always right. This is why understanding how the code works is important. Getting to a place where AI can supplement and speed up your coding process will be much more valuable than solely relying on AI to write the code.

After expanding to view the full result (see Figure 1.2 and Figure 1.3), you may be thinking, “Why did I pay for this book if I can just Google the results?” It’s a fair question, but here are a few things to consider:

• These results have no idea about your data (see our discussion of prompting in Section 1.2.2).

• If you don’t already have the underlying knowledge of a random forest model and how it works, your ability to correctly edit it is limited at best.

Figure 1.2     Google Gemini AI Results for Building a Random Forest Model (Part 1)

Figure 1.3     Google Gemini AI Results for Building a Random Forest Model (Part 2)

So, while Google’s AI result may seem extremely convenient, it should be considered with caution. As a rule of thumb, avoid blindly copying and pasting this code, as it almost always requires edits.

Python Library Documentation

You’ll often see the actual Python library documentation in your search results (the first non-AI summary result in Figure 1.1). The quality of documentation can vary, and it can be dense at times, but the libraries leveraged in this book all contain robust and high-quality documentation. The packages that don’t have the best documentation are often the ones with less usage, which usually means they have niche applications. For example, network analysis in Python doesn’t have an established, go-to library like data manipulation has with pandas.

The developers and maintainers of the library usually provide simple examples of how to use each function. These no-frills examples are especially helpful when you’re using new functions or packages. As you become more familiar with each library’s documentation, you’ll find that many of the other sites that appear in your Google search simply use the library’s documentation examples. This is why we recommend starting with the library’s documentation.

What Is a Python Library?

The concept of a Python library may seem very abstract. Python is considered a high-level programming language, which means you don’t need to understand how the core Python language works to be able to code with it.

Python libraries are what make Python accessible to individuals who aren’t computer programmers. The libraries are built by writing Python code and packaging it for others to use. It cannot be overstated how much time this saves you when building a machine learning model.

Stack Overflow

You’ll notice that Stack Overflow, a community-based platform where users can post questions and answers, isn’t in our search view—it was six additional sites lower in the results. Stack Overflow seems to rank higher for queries with more specialized use cases. The nature of this platform invites specific questions, likely explaining why Stack Overflow isn’t showing up in our top search results.

GeeksforGeeks

GeeksforGeeks, a personal favorite, has been one of the most helpful resources. The platform is article-based, so it’s a middle ground between more specific questions (Stack Overflow) and really general information (library documentation). The biggest limitation is their concise, blog-post-like format, so you get more polish and clarity at the expense of content.

Reddit

Reddit is a place to find really niche answers. However, my success rate at finding the answers I’m looking for in the machine learning space is relatively low. If you’re a general Reddit user, I promise this isn’t a dig on the entire platform. (I’m aware of the passion of the Reddit user base!)

Paid Sources

There are a number of paid sources that you’ll likely come across. One example you see in the search results in Figure 1.1 is DataCamp. They have some free resources available at the time of writing, but the bulk of their content is behind a paywall. I personally love DataCamp and recommend it to people who are getting started in programming. However, their content leans into the category of course learning versus ad hoc research and a source for questions.

Towards Data Science and Medium are blogs with great content behind a paywall. In my experience, though, there isn’t a noticeable difference between GeeksforGeeks’ content and these two paid sources.

Specificity of Python Resources to Answer Questions

It can be helpful to consider these resources based on where they fall on a scale from highly generalized (1) to highly specific (5):

1. Library documentation

2. AI summary

3. GeeksforGeeks

4. Stack Overflow

5. Reddit

1.2.2    Prompting

The speed at which generative AI progresses is enough to make your head spin, so this section of the book will probably age poorly very quickly. I wrote the content for this book over the course of 2025. At the beginning of my writing process, many companies were still working through the security implications of enabling their workforces with generative AI, and the conversational interface was the main way to interact with ChatGPT. By the time I was revising my final draft, OpenAI had released their own web browser, and most companies were well on their way to integrating AI into their processes. The impact of LLMs and their ability to assist you in writing your code has a lot of potential. Throughout the book, you’ll see examples of how I use generative AI when writing code and working through machine learning problems. Here are some considerations for leveraging generative AI tools when working on machine learning...