infraSOFT Machine Learning module for Data Cleaning and Forecasting
Machine Learning is used in many industries when a decision has to be made that requires analyzing a large amount of data. For example, what will be the likelihood of failure next year, and in 5 years, of a 6" CI pipe, installed in 1927, located in zip code 11111 where soil is mixed landfill, that has never broken before? What is the most likely material of a pipe based on its location, year of installation, and diameter? Machine Learning relies on our current access to computing power that allows creating and testing the many (and when we say many we mean many, millions, billions) possible connective paths between all available variables analyzed through a variety of models, and settles for the most likely connections.
infraSOFT.ml, the Machine Learning module of infraSOFT, generates a likelihood of failure score for each pipe and each year in the future which allows ranking pipes. That's not all; we also generate the aging curve (break rate by age) which we use to determine the Effective Useful Life. infraSOFT.ml also helps filling in missing values.
infraPLAN has modeled pipes and breaks for decades. This has given us unparalleled experience to develop high-performing Machine Learning models thanks to:
All models are validated by comparing predicted and actual values.
Note that as powerful as Machine Learning can be, the above tasks do not end when the likelihood of failure, aging curves or Effective Useful Lives are generated (that's the easy part!). We then make sure forecasting results and rehabilitation plans can be interpreted thanks to the STATS module, and that the underlying cause of data issues are well understood and remedied (CLEAN module); then we make use of all those results to generate an optimal plan of rehabilitation (PLAN module); infraSOFT.ml and failure forecasting are just one tool in our box; infraSOFT's has many more.
Copyright © 2018 infraPLAN - All Rights Reserved