How Applied Machine Learning is Driving Team Focus & Reinventing Human Resources
Highlights from the November 2018 Bots & AI Meetup: “Applying Machine Learning to Drive Team Focus & Insights”
On November 12th 2018, The Bots + AI Meetup had its November event exploring machine learning in team productivity for project estimation and HR/engagement. This event provided interesting contrasting use cases to the recent excitement over deep learning based algorithms that perform well with huge amounts of data but can be difficult to explain and justify as well as potential model instability.
Driving Team Focus
In the first use-case for Hive, agile team performance and project delivery prediction, the input data is highly limited as you can only collect data after each task is completed. Getting thousands of data points is unrealistic for most projects and would take too long even for those before effective predictions are available.
Additionally, development is complex and variability/risk is a concern so a single prediction of a delivery date is far less useful than a confidence interval. The Markov Chain Monte Carlo algorithm is much better suited to this purpose.
Reinventing Human Resources
HR is another area where insights in engagement, retention, and hiring need to be accepted and approved by managers and be explainable.
Again, regression based models were favored over deep learning approaches as decisions affecting personnel needed a high degree of explainability to the HR professionals responsible for their implementation.
Clustering and extensive visualization in charts and connection graphs were key enablers in understanding. Topics ranging creating predictions on cities and schools to recruit from to including the value of a candidates network for future hiring were explored.
Gramener also offered their own clustering algorithm, Group Means, with a live example viewable in the resources section below:
Cluster Analysis using Group Means Visualization
More tables had to be brought in the room for adhoc seating due to the overwhelming response for the image recognition workshop by IBM Developer Advocates Helen Lam and Nicholas Bourdakos. Helen got the room started thinking of common as well as creative image recognition use cases including very recent use cases of images from drones or satellites to track forest fires.
Nicholas Bourdakos continued getting the attendees setup with the open source Tensorflow framework and getting their models working. Attendees left with a working image recognition deep learning example in Python on their laptops as well as experiencing IBM Watson Studio for a training GUI experience.
Apple’s partnership with IBM brings exciting opportunities for machine learning on the edge computing arena. Models can run on smartphones without needing internet or cloud access leading to enhanced privacy and access possibilities. The workshop also went through how the model being worked on could be exported to CoreML and run on iOS devices.
Attendees that finished early explored some of the other demos including a drone forest fire exploration.
IBM Workshop Resources
Original info on the IBM Machine Learning Workship can be viewed here.
