AI and Machine Learning as Analytics, let’s simplify!
What do analysts do with data? You would think the title “analyst” is self-explanatory – they just analyse data – but the reality is far from “just analysis”. An analyst or a data scientist takes the data through a journey from raw incomprehensible numbers to valuable insights that a layman is possibly oblivious to. This process includes acquiring the data, loading it to suitable environments, cleaning and transforming the data into a useful format, and then generating insights from it. Just the last step can be a wide spectrum of tasks which could be anywhere from simple reporting/visualization to advanced modelling for predictive analysis.
So where does advanced analytics come into the picture? Where do we get to use Artificial Intelligence and Machine Learning to extract the power from data?
Bring out the buzzwords!
We hear a lot of fancy words thrown around data analytics. Companies say words like Machine Learning, AI, Predictive Modelling, Deep Learning, Big Data, IoT etc when they want to talk about the “next generation work” they are doing. Yes, these are the common buzzwords, but what do they mean and why should you be interested in them?
Through machine learning and statistics, we are able to decipher and realise the potential information that data holds. From predictions about the stock markets to analysing clusters of people from their online shopping history, Machine Learning has given us capabilities to find patterns in data quite easily – now more than ever before. AI can be a part of the “generating insights” phase of a data analyst’s pipeline. From a broader perspective, Machine Learning can be categorised into two major groups – Supervised and Unsupervised learning. Unsupervised learning usually includes techniques like clustering (or grouping datapoints), where there are no “target variables”, or pre-trained data to learn from. A Data Scientist looks at patterns and intricacies in the data without knowing what they might look like. It is often quite difficult to imagine data in high dimensions, but with such methods we can understand how each variable in our data interacts with another.
On the other hand, Supervised Learning is about correlating behaviours, or variables in the data to a “target class” or a prediction variable. This is what the majority of an analyst’s work would look like when it comes to machine learning. Predictive analytics is the most common subset of AI that companies and individuals work on. We look at possible causes that lead to changes in our target variable and predict future data to fall into one or more of the target classes. Predictive analytics could be used in most major sectors of the economy. Retail, sports, health, weather, energy, banking, and even social media/internet data can benefit from supervised learning. Whether it is improving customer satisfaction in retail, or predicting delays in airline travel, or forecasting the stock market around global events – Predictive analytics looks at past patterns and speculates future behaviour of data. Paul van Loon, Head of Analytics at Forecast, speaks with Atif Hussain, Managing Consultant at Fyte, about this:
There’s a lot more to data analytics than Machine Learning!
Is a Data Scientist working on machine learning techniques all the time? No, there’s a lot more to Data Analytics than just Machine Learning or AI. At Forecast, the analysts don’t just bring out the valuable insights from data, they also tailor these insights to the audience they’re presenting to. It could be someone from a finance background, or someone from marketing – Forecast makes sure that the message is delivered in a way anyone can understand. Data analysts and scientists use visualisations and summary presentations that show the gist of the analysis, capturing the KPIs of what the data shows.
In summary, we see that a data analyst or scientist usually works through the entire pipeline, from getting data to analysing it to delivering a clear message to the audience. There may or may not be Machine Learning techniques used in this pipeline, but there surely is a lot you can do with any meaningful data.
