Specific methods relevant to data analytics

The main data analytics methods include

• Machine Learning
• Text mining
• Sentiment analysis
• Systematic reviews
• General statistical analysis
• Visualisation

Let me go through them, one by one. Machine learning, as the name suggests, aims to mimic the way we humans are learning. We might observe patterns, shapes, and trends in a given data set. However, we usually are only able to consider small datasets. So, how can we scale up? Can we create methods and frameworks so that the same procedures are carried out automatically?

This is the aim of machine learning. There are two types of approaches to ML. The first one is supervised learning. Based on a labelled dataset (where the expected output related to the corresponding input is given), the learning process is facilitated. In other words, we ‘educate’ our model.

On the other hand, unsupervised ML does not have any labelled data. It will learn as it progresses through the computation. Despite the differences, their objective is identical, that is finding data trends, and/or classifying data. The main ML models fall into the following:

• Regression: a trend (usually depicted by a ‘curve’) is identified so that a prediction can be carried out
• Classification/clustering: given a dataset, this model will group them into sub-sets based on specific attributes or properties
• Dimensionality reduction: attempt to remove uninformative dimensions from a multi-dimensional dataset

We communicate via spoken and written languages. The amount of information shared, stored, and multiplied across texts, books, novels, etc, throughout human history is staggering. Text mining and sentiment analysis are two ML approaches to analyse textual sources.

More specifically, the former aims to extract concepts, relationships, entities, and semantic information, whereas the latter, despite being part of text mining, focuses not only on what we are talking about but how we are talking about it. In fact, often it is not just important to identify linguistic concepts, but the corresponding opinion held by individuals. This is particularly relevant in, for example, marketing, where sentiment analysis is used to analyse how people discuss brands.