In a world that is inundated with data, introducing Machine Learning can help companies stay ahead of the curve by boosting efficiency and improving their decision-making.
In this blog, we will dive beyond the buzzwords and explore the concepts that underpin machine learning.
Machine learning is a subfield of Artificial Intelligence (AI) that involves the development of algorithms and models. These algorithms provide computers with the ability to learn for themselves through data, rather than being explicitly programmed.
Here, we attempt to put in a nutshell, the end-to-end process of building an ML model.
- Understanding the Problem
Before anything else, make sure you have a clear definition of the expected outcome.
Some questions to ask yourself to help define a solution are as follows:
- What are you trying to model?
- Are there any ethical considerations surrounding the problem? For example, are there any biases present in the training data?
- Can an incorrect prediction have a significant negative impact? If so, should any predictions be made?
- Training time vs prediction time? Do you need a model with a fast prediction time, but the training time is not of huge importance? An example of this could be a self-driving car.
- How important is explainability? If the model makes accurate predictions, but you are unable to explain how exactly it is doing so, does it still satisfy the aim of the problem?
The answers to all of these questions should help guide your choice of which machine learning models are suitable.
- Understanding the Data Available
The phrase “garbage in, garbage out” is the mantra of machine learning.
Spending time cleaning and transforming your raw data into a suitable format is a crucial step in the process.
- Types of Machine Learning
Once the goal and the data available has been established, the ML models that are applicable will have been significantly narrowed down.
There are two common types of Machine Learning; Supervised Learning and Unsupervised Learning which work on different types of data and solve different types of challenges.
- Supervised learning: A model is trained on a labelled dataset, where the input data is paired with the corresponding outputs.
- Unsupervised Learning: A model is given unlabelled data and can uncover hidden patterns within the data.
- Common Algorithms
Depending on the problem, here are a few examples of common ML techniques:
| Supervised Learning |
|---|
| Linear Regression |
| Decision Trees |
| Random Forest |
| Support Vector Machines (SVM) |
| Unsupervised Learning |
|---|
| K-Means Clustering |
| Hierarchical Clustering |
| Principal Component Analysis |
- Model Evaluation
Once you have selected a few candidate models, the next step is to fit them to the data and measure the performances.
One way to evaluate the performance of a model is by using Cross-validation. Cross-validation uses different portions of the data to test and train a model on different iterations, and the results are combined at the end, giving an idea of how well the model would perform on unseen data.
It is important in the validating and testing stage to identify whether the model is overfitting or underfitting the data. Overfitting is where the model learns the training data too well (including noise in the training data), and performs poorly when generalizing to unseen data. Underfitting is where a model is too simple and is unable to capture underlying patterns in the training data (for example, using a linear model when the relationship is not a linear one).
Hyperparameter Tuning is an important consideration when trying to boost model performance. Each algorithm has it’s own set of hyperparameters, and hyperparameter tuning involves finding the combination that optimises the model’s performance.
Finally
Machine learning is a powerful field, and keeping up to date in the latest advancements and best practices is crucial for practitioners. Additionally, understanding the specific needs and challenges of the problem at hand is essential for selecting the most appropriate machine learning technique.