EXPLORING THE MYSTERIES OF AUTOMATED AI — THE FUTURE OF DATA SCIENCE!!
Modern Computers have built the endless ocean of data in the past few decades and have made the data-driven decision making inevitable in every industry. The vast stream of data has increased the importance of Data Science in the world. Machine Learning lies within the scope of Data Science and has gained popularity due to its ability to extrapolate the historic results in making future predictions and taking decisions accordingly. However, the lengthy iterative processes essential for making Machine Learning productive often slows down the decision making process.
On the other side, the IBM Watson Studio AutoAI is believed to automate the Machine Learning experimentation through automated Rapid Prototyping. Is it really effective in performing the Machine Learning Experiments better than the manual iterations by Data Scientists?
In an attempt to find the answer to this question, we have tested three types of predictions in IBM AutoAI i.e. for i- Binary Classification ii- Multi-Class Classification and iii- Regression Analysis. The key parameters of the datasets, business case of predictions and the Results of Machine Learning Experiments through AutoAI have been elaborated below:
1- PREDICTING THE JOB SEEKING MINDSETS OF DATA SCIENTISTS (BINARY CLASSIFICATION)
In order to test the Binary Classification (Yes/No) capability of the IBM AutoAI, the AI Experiment of analyzing the mindset of the Data Science Professionals (who enrolled in a particular University) conducted by us to see the prediction capacity of AutoAI in analyzing whether the course participants are planning to switch their existing jobs OR just taking the course to improve their current job performance? The parameters given for this prediction include information about participants’ residence, their Qualification and Experience and the type of Enrollment in University courses. So, can the AutoAI Model point out the persons who are looking for a new job from such basic data?
The IBM AutoAI recognized the prediction type to be Binary Classification & suggested the suitable Models/ Algorithms automatically. The experiment is automatically conducted by the IBM AutoAI and 8 Pipelines are generated by default. Upon conducting the automated AI Experiment without any manual intervention, top performing Pipelines yielded about 79% optimized accuracy i.e. it successfully identified 4 out of 5 Data Scientists who were looking for a job change. This IBM Automated AI Experiment is, thus, proven to be successful in identifying the Job Change Mindset; and it’s feasible for the companies interested in finding the talent pool of Data Scientists looking for new job opportunities on one hand; and measuring the percentage of the Data Scientists who are satisfied by their current job roles on the other hand.