Artificial intelligence, machine learning and analytics are the buzzwords. If you want to start specializing in Machine learning through MBA in AI and ML or otherwise, in your managerial career knowledge of these terms is going to be vital. With this knowledge, one will be able to separate the real value of these technologies from the hype.
In the curriculum of distance MBA in AI and ML, in depth coverage would be provided for analytics techniques, methods and tools. Here let us try to understand basic terminology used in different types of data analytics with examples.
Type of Analytics
Data analytics is categorized into mainly three types.
|Descriptive Analytics||What has happened?|
|Predictive Analytics||What will happen?|
|Prescriptive Analytics||What should be done if this happens?|
Descriptive analytics the most basic form of analytics. When we have past data, we want to know what worked in our favour and what did not. This is a postmortem report. Using various data aggregate and mathematical functions one can know what was average, mean, standard deviation etc. for the given data set. It tells what happened in the past and the knowledge or insights or patterns generated from this data can help plan in future.
Most of the businesses who use databases and data warehouse use this kind of analysis as part of their continuous improvement in business decisions making or quality.
For example, you have been spending every month some dollars on various marketing initiatives and tracking the lead generated from those. If we collect the daily data for a year, we can find out which initiative gave maximum results, what was the average cost per lead, which mediums were completely ineffective for some set of customers etc. and this would help us plan and tweak the marketing plan in future.
Predictive analytics in new entrant in this space. Here the past data is captured and used as training data for the software. The software algorithm takes this input data and creates a statistical model. This model is built using probability theory. So if you have got success with certain combination of input parameters, what is the likelihood of future outcome be a success can be predicted using the model. As one can add different input parameter which are contextually relevant and then a correlation can be found out between input and output.
Let us understand more about predictive analytics with an example. There is a company which manufactures fertilizers. It is believed that the sales of fertilizers is linked with the amount of rainfall. We have a lot of past data about average rainfall in the state/region during the week and month. We want to use this data to predict what would be the sale of fertilizers.
Let us see how to model this using two variables X and Y. Here X is called independent variable and Y is called dependent variable. In our example, X is rainfall which determines dependent variable Y which is sale of fertilizer. Using the available data, we try to find out a mapping function from the input variable X to the output variable Y.
Y = f(X)
The input data or values of X(rainfall) which determines function f(X) is called as training data set. This training data set is fed to a predictive algorithm called “linear regression”. This algorithm creates a mapping function f which is called the predictive model and is used for forecasting of Y based on new values of X.
This method is also called as supervised learning because it resembles an instructor observing and helping the learning process under his supervision. In supervised learning methodology, the predictive algorithm goes through multiple iterations till desired level of accuracy and precision is achieved.
There are many algorithms like regression, random forest for classification, support vector machines etc. are used in predictive analytics. Predictive analytics is one of the key aspect in learning data analytics for distance MBA in AI and ML.
Prescriptive Analytics is the next step beyond prediction. It advises the business on what to do to increase probability of success.
Based on past data of various parameters, prediction can be done to focus on customers with particular qualities. A bulk email blast to these is done and some customers start showing interest by clicking links embedded in the email or visiting your website.
Here, every lead is at a different level and probability considering the readiness for purchase. To effectively handle this there is concept of sales pipeline in sales and marketing. Various leads when they are tracked are scored based on their probability of conversion. There can be different actions proposed for leads at different stages. The leads who show lot of interest in going through the website till the shopping cart, may get converted as buyers if sales team calls them immediately. The other leads which have not yet decided and are not going beyond your product page, can be sent an email of more information or sent a discount offer to move them towards closing the deal.
Any candidate who is pursuing distance MBA in AI and ML would come across these types of analytics in his curriculum with relevant case studies.