Predictive Analytics is a valuable process to solve real world problems in many industries from health to business. My journey into learning Big Data looks similar to the picture you see here, scary. It’s somewhat comical that’s why Hilary Clinton should think twice before messing with the servers. Too bad her cheating ways didn’t get her no wheres. Alright, back to the subject. I’ve seen how beneficial predictive analytics has been with proof from so many companies using it, now to just get down to the nuts and bolts of things. In this article, I’m going to explain the process of putting it all together, the very beginning before you get to the solution.
The 6 Step Process of Predictive Analytics
1.Outline the Project
First, you need to outline the project solutions, scope, deliverables, data sets and the business objectives. This is usually the project manager’s responsibility with the business intelligence involvement in the very beginning of the procedure.
According to author Ali Rahim, from his article Best Practices for Business Intelligence and Predictive Analytics, many aspects are involved in this development, such as business objectives, background information, prior history resources, modeling objectives, constraints, assumptions, risk, contingency, tools, techniques, criteria for success, terminology and project plan.
Here’s an example of the breakdown of the Project:
2. Assessing the Data
This consists of using data extraction, data mining, exploring data resources, retrieving quality data and verifying the data to ensure the desired outcome will be achieved. There are multiple people or teams involved sometimes. From observing the breakdown from the above graph, we can see that the BI Developer and Statistician is involved mostly with assessing the data.
3. Preparing the Data
This is where careful examination is done. It usually involves selecting the data fields, data cleansing, and constructing the data searching for missing fields or values, merging other data, formatting data, integrating other data sets, then doing a final report of a data set.
4. Data Modeling
This step involves having a data model or technique in mind. It’s a great idea if you ask a data mining expert that could give you some recommended tips for using the right model for your project. Make sure to document and test everything. Build a test model and have several models in case something fails or succeeds. For example, recording a linear regression for sales target or using multiple regression for doing cost analysis for a warehouse.
5. Evaluating the Model
Next, is to evaluate the model against the required criteria and business goals. Continually ask yourself: Does this model meet the goals of the desired outcome? Does the model fit the product? Be sure to validate the predictive analysis models before deciding to deploy the project. You want to make sure everything works out and don’t assume that one model fits all cases. In other words, all companies are different and every project isn’t necessarily the same. Also, most importantly document all information to avoid any errors, I cannot stress this enough.
Finally, in this last step is the icing on the cake. This involves creating the UI application design , workflow design, platform, more development and more testing. Keep in mind there is no perfect model for everything. Test and keep testing.
I’m just learning the ropes of analyzing data. Predictive analytics is fascinating and it’s not always perfect just like us. We live in a big data and it is a small world after all. I’ve yet to do more exploring and I intend to dig deeper to learn about modeling. My mission is to help people be successful and make the world a better place. Join me next time through My Journey into the Data Analytic World: Part 2 Data Models.
Any thoughts? Are you perhaps an expert in predictive analytics? Share some tips with me. What do you think about my Hilary Clinton matrix picture I found?