Release 0.0.5 update

[Kevin Fasusi]

The exponential smoothing forecasts for each SKU, using evolutionary algorithms for optimised alpha and gamma values, and load scripts for the reporting suite were previously computed sequentially. The previous implementation resulted in long response times for hundreds of SKUs. For an industrial application, this is obviously prohibitive in use cases where the inventory holds tens of thousands or even hundreds of thousands of individual SKUs.  Good news, release 0.0.5 (when released) will have overcome this particular problem. Both the forecasts, the load scripts and the Monte Carlo simulation, now take advantage of the  'concurrent.futures' and 'Multiprocessing' modules. The faster implementation makes use of multiple cores on a system. Release 0.0.5 will move the library in the right direction for scale. Tests on a hexacore workstation with 12 threads have been excellent. There will be more tests with much bigger systems using a cloud platform most likely AWS.  The  GitHub wiki will host the system profiling and timings around the time of release. 

A primary focus, when the time has permitted, has been putting the reporting suite together. The last application page added in the up and coming release is the Simulation view. This view allows the user to run the discrete event simulation from the browser and retrieve the summarised results for each SKU. The number of runs can be specified, and the user can search the results. The plan is to add more charting to this view. After the graphs and charts have been added to this view, the reports for inventory and demand planning will be considered feature complete. Although, there will be efforts to make the results persist in a database and more charts added after implementing more forecasting methods.  In regards to the simulation view, at the moment the retrieved summaries do not get stored. The expectation is that for critical use the simulation should be run from an interactive shell or IDE. 

After release 0.0.5 we will redirect development focus to the missing parts of the core library including ARIMA forecasts, production and distribution planning, routeing and scheduling as well as warehouse optimisation and financials. 

As mentioned in a previous post, there are plans for a formal roadmap. Cleaning up the documentation and making clear the entry points for contributions is on the ToDo list. Using the Github wiki for more immediate communications about developments is part of this plan of action. As always if there is anything you want to share post it here or on the GitHub issues board, any feedback is useful. Have a good one!