The Future of Machine Learning

Moving data analysis to the cloud is only one part of the way machine learning projects have changed over the last few years. Since the benefits of adding automation, Artificial Intelligence (AI), and machine learning to many different parts of business operations are now clear and don't need further proof, companies are now focused on more permanent solutions. In fact, the natural follow-up is to think about finished products that can complete the full data cycle, from data collection to visualization.

There are many ways to create data analysis flows that can consume data as it is created and return results and visualizations after applying machine learning models. Cloud services make this task easier and more efficient.

Automatic machine learning is the current tendency in data analysis, where several machine learning models can be tested...

Automatic data analysis flows

A few years before this book was written, businesses were approaching machine learning with the phrase Let's see what this thing can do... in mind. That is not the case any more. The value of using analytics, machine learning models, AI, and advanced visualizations to understand, simplify, and predict the outcome of many different situations is clear. This value is measured in terms of money, time, and effort savings, which leads to better and faster business decisions.

As a summary of what we have learned throughout this book, we can list the different parts of a data analysis flow:

• Data collection, usually from diverse sources
• Data cleansing and preparation, including exploratory visualizations
• Choosing a machine learning model that suits our data
• Training the model with historical data (if we are talking about supervised learning)
• Mining...

Automated machine learning

There are several tasks that are crucial for the success of a machine learning model when applied to solve a given business problem, for example:

• Data pre-processing
• Feature engineering
• Model selection
• Optimization of the model hyperparameters
• Analysis of the model results

These tasks were usually performed more or less manually by experts in the field. In recent years, there has been a growing interest in democratizing machine learning, allowing for non-experts (sometimes called citizen data scientists) to use, improve, and apply machine learning to concrete problems. Automated Machine Learning (AutoML) targets that specific need.

In general, the building process of a new model can be described as in the following diagram:

Following is the process for building of new model:

• Input data is pre-processed and used to build the best model features
• Based...

Summary

The last chapter of the book is thought of both as a summary of all chapters and as a window to what can be done beyond Excel and in the future. Automated data flows and machine learning model generation simplify analysts' work and speed up the decision-making process.

Hopefully, you now have a wide view of what machine learning is, how to use it in each line of business, and what are the most advanced alternatives to what was known before reading this book.

Questions

1. What part of the analysis flow is different for supervised and unsupervised learning?
2. Why is data cleansing a continuous cycle?
3. Explain briefly what model hyperparameters are.
4. Which steps can be performed automatically by AutoML?

Further reading

• Azure Machine Learning Service Documentation: https://docs.microsoft.com/en-us/azure/machine-learning/service/
• Machine Learning on AWS: https://aws.amazon.com/machine-learning/
• AI and machine learning products: https://cloud.google.com/products/ai/
• Announcing automated ML capability in Azure Machine Learning: https://azure.microsoft.com/en-us/blog/announcing-automated-ml-capability-in-azure-machine-learning/
• H2O.ai: https://www.h2o.ai/
• Cloud AutoMLB: https://cloud.google.com/automl/
• AUTOML: http://www.ml4aad.org/automl/