Big Data Fact:
The Buzzwords in Big Data Analytics
“Artificial Intelligence”, “Machine Learning” and “Neural Network” are certainly hot topics in big data analytics in this era thanks to the expanding data repositories, advancements in computing algorithms, enhancements in data processing capacity and of course better affordability in data server space.
Image credit: https://anyip.io
Around the world, millions or perhaps even billions of dollars have been invested in Artificial Intelligence and Machine Learning to resolve everyday challenges such as improving the conversion rates of advertising. The customised online ads or the “recommended searches” that we receive on our newsfeed is one of the key results from these investments.
“Don’t they all mean the same thing?”
No, they are not quite the same thing. Let’s start off with a few quick definitions
Artificial Intelligence (AI) – AI enables machines or devices to dynamically self-learn from its past experiences and refine the algorithms based on new data it has acquired. AI arms machines with the ability to “think” and empowers them to perform natural tasks and display reactions like humans do.
The introduction of early AI could actually be seen in our lives way back than some of us can remember. For instance when IBM introduced the chess playing program “Deep Blue” which the reigning chess champion Kasparov succumbed to decades ago. In more recent times, we have also seen many other examples of AI in our lives; from weak AI smart home devices like Google Home and Amazon Echo to advanced AI like Sophie the robot citizen. Just looking at the progress of AI in the world, it almost seems like Tony Stark’s Jarvis may be within our reach soon.
Machine Learning (ML) – ML is one of the more popular methods in achieving AI and one of the advanced methods used in big data analytics. It is an extremely useful tool in refining task specific algorithms; ML involves loading the machine or processor with large networks of data and parsing all of it, recognising patterns to adjust the accuracy of the algorithm. With ML, systems can be “trained” to classify, predict and forecast. For instance, how our digital cameras recognise faces or scenes or how the agents in ‘Criminal Minds’ match their suspects in their database are classic examples of ML in practice.
Neural Network (NN)– A NN is a mathematical model of interconnected elements (used to predict outcomes) which resembles the functions of a neuron in the human brain and how it reacts to stimuli. While there are many who think that NN and ML refer to the same thing, it is actually just one of the techniques in parsing data that is commonly used in ML. The actual parsing of data in NN involves utilising a complex system of weights assigned to each element forming clusters.
How do they relate to my business?
No matter the size of your organisation, you need data insights to evolve your business. Broadly speaking, insights can be catagorised into 4 key types:
1) “Descriptive” which allows you storyboard your data, giving it meaning. For instance, a descriptive insight could tell you that there was a 50% increase in repeated patronage last month.
2) “Diagnostics” which uncovers the reasons for a phenomenon. e.g. insights on why was there a 50% increase in repeated patronage last month so that you an emulate it again next month.
3) “Predictive” that shows the relationships between variables in your data through advanced analytics such as data mining, regression modeling, neural networks, machine learning or even artificial intelligence. These insights provide reliable forecasts of what may happen in the future such as predicting your customer’s reactions to a stimulus.
4) “Prescriptive” insights are obtained through a relatively new form of analytics. Essentially, these insights are uncovered through various advanced statistics, models and tools to prescribe a variety of possible actions which you could take. It quantifies the effects of possible decisions through probability allowing you to select the most ideal action.
All 4 types of insights can affect any business significantly. These insights complement each other very much and are best acquired in a progressive manner. Afterall, “Predictive” and “Prescriptive” insights usually stem from “Descriptive” and “Diagnostics” type of insights.
Do I require advanced analytics IN my business now?
The good news is that you have been collecting data since the day which you have started your business: data from your end-users’ feedback, customers’ details, transaction volumes, eCommerce platforms, internet-of-things, competitive intelligence, social media, supply chain, product and service experience. These data can be used to drive the insights which you require. However, the various data sources need to be integrated for advanced analytics to commence effectively.
Any insight is only as good of a quality as the data that drives it. To determine whether you should be utilising advanced analytics in your business, it is usually a wiser move to first take a look at the quality of data used to drive your insights.
5 questions to ask before Jumping the Gun
As enticing and useful as they sound, advanced analytics may not necessarily be required in your business right now. Instead of focusing on the tools or getting buried by the analytics right away, think about what they are meant to support. Ask yourself these 5 questions about the data you have or are collecting before you decide.
1) What are you intending to use the data for?
We have encountered businesses that jump straight into research or analytics without first identifying the research problem or constructing a proper hypothesis (a hypothesis is a statement, not a question); while a few others have rather generic objectives. This does not only waste your resources but also lower the effectiveness of your insights.
2) What type of data are you collecting?
Different types of data enable different types of analytics. If it is a dipstick on customers’ sentiments that you are after, there isn’t a need to go “hulk” on NN or ML right now. You do not require advanced analytics to tell you that a decline in satisfaction implies that you need to take immediate action.
But for instance, if your intent is to crawl the social media for a sentiment analysis, perhaps ML would be worth a thought in facilitating the classification of comments on social media for your quick takeaways.
3) Do you have enough data?
The margin of error from your analytics declines as the quantity of data increases. To experience a higher accuracy in your prediction models, you will need access to large quantities of data. Unless there is sufficient data to mine, we can get to that eventually as the size of the server increases.
4) Is your data representative and reliable?
Have you got all your grounds covered? Only collecting data from selected data points gives you a myopic view. For instance, data from your website gives you only one side of the coin. But when you integrate data from the various sources, you will enjoy a panoramic view.
Also many consumers are passive in nature – not all would proactively share a comment or review your products / services. Still, we simply can’t ignore them entirely as it may result in false positives or negatives on how your customers truly feel.
5) What type of insights does your business need the most right now?
Certain techniques of analysis have accompanying assumptions and can only be applied when the conditions have been fulfilled. Analyse, identify and stay focused on the type of insights which your business requires most right now.
Don’t do it If You Can’t follow It
The relevance of data science and advanced analytics in businesses have rapidly grown in dominance over the last decade. Like many others, if you are contemplating on getting some action yourself to gain that edge over your competition, it is surely the right time to do so in our current data-driven world. Still, it is important to understand that the outcomes in predictive models may not necessarily be intuitive and may sometimes require you to take that leap of faith. Afterall, the difference between success and failure is not about our tools, abilities or ideas but about mustering the courage to take the calculated risk.
Why Not Say Hello?
Or do you have other questions in mind? We would ♥ to help!
