You are currently viewing Getting Started with Machine Learning

Getting Started with Machine Learning

Starting out with machine learning (ML)? You will typically wonder if the time and effort spent learning about this technology is worth it. And if it is, then how do you get started and where are the pitfalls of this. Will machine learning capability be useful to businesses?

Machine Learning 101

ML is definitely worth it, it is the future and will be pervasive throughout our lives. Intelligent machines will predict what will happen, explain things, automate a lot of tasks from driving cars to office work, and help business leaders in making optimal decisions by crunching the data for them. Even software professionals will need to enhance their skills because ML will be used to develop systems, reducing the need for traditional software development. 

Getting started on ML might seem daunting. But you don’t have to enroll in a course, as most are designed to train data scientists. You might instead just want to use ML as a casual practitioner, not an expert, just to familiarize with it.

So the best way to get a handle on ML would be to read up on how it can help you with your goals. Collect limited training data to train the ML and sign on to an ML platform that talks you through the process of building models to predict. Many prefer open-source ML platforms, but these assume expertise in programming, in coding with Python, and this might deter beginners from starting out in ML.

As you research ML platforms, be sure to go for user-friendly ones with easy-to-follow training videos and generous support. This will help you with any issues you might have. And when you have your first ML model, be sure to compare its predictions with what you actually observe. After all, ML is a process that entails a lot of refinement over various iterations.

Here are some examples:

Are you a subject matter expert on the operations process of an enterprise? This operations process will need data from various sources, including the cloud, and decides how to execute. This will require writing rules on how to cover various possibilities, which in turn will include many exceptions. Investigating each request can be exhausting. You might have wondered if you can build a system to help you out. A system that will suggest how to process the request and the reasons for it. And you have heard about ML and how it can help with devising such a system and thus lessen your workload. But the main hurdle is that you don’t know how to get started, how to invent the training data for the machine learning platform. 

Here’s another example. If you are an economics major collecting tons of data for your research project, you would like insights on what this data reveals about the outcomes you are analyzing. How should you do this? By studying correlations between variables? You know about ML and how it could help. But again you just don’t know how to get started with learning ML in order to use it for your project.

The hurdles of accessing and using ML may be what’s holding it back from widespread adoption. So, how can these hurdles be overcome? 

Training Data

Training data is a major challenge. There are many training data providers who provide crowdsourced training data that have been sanitized as best as possible to free them of biases. Like Lionbridge and Appen. But on your end you might not be able to divulge your process details as they are confidential. Or you are exploring ML on your own but do not have a budget to pay for these services.

So when researching ML platforms, look for those with “connectors” to the most common sources of data, including your systems and the cloud, that can automatically extract information elements you are interested in. These connectors should be able to:

  • Get information elements automatically.
  • Assemble and file them in appropriate order.
  • And go on training statistical models from the training data in the file so you can begin the predicting process.

These connectors should be able to periodically pull in data, once a day or regularly, and collect enough over days and weeks so you can get the data you need to start with ML.

Now, how do you use this data to understand ML and build models? Make it easy on yourself by looking for an ML platform that automatically builds models from training data. Ideally it should explain the various steps involved. So you can understand how the platform is processing the information, the statistical methods it is using, and how it selects the best models to learn and predict for your purposes. Without these answers, you won’t be able to fully understand the process. And you will have a hard time discussing your work. 

So, to sum it up: You have to do ML to learn it. But machine learning has already reached the point where it is accessible. Now you can stop worrying about how to get started.