US 12,271,917 B2
Skills and tasks demand forecasting
Subhro Das, Cambridge, MA (US); Wyatt Clarke, Bronxville, NY (US); Sebastian Steffen, Ithaca, NY (US); Prabhat Maddikunta Reddy, Danbury, CT (US); Erik Brynjolfsson, Palo Alto, CA (US); and Martin Fleming, Glen Ellyn, IL (US)
Assigned to International Business Machines Corporation, Armonk, NY (US); and MASSACHUSETTS INSTITUTE OF TECHNOLOGY, Cambridge, MA (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Jan. 27, 2021, as Appl. No. 17/159,449.
Prior Publication US 2022/0237635 A1, Jul. 28, 2022
Int. Cl. G06Q 30/00 (2023.01); G06F 9/445 (2018.01); G06F 40/40 (2020.01); G06N 3/044 (2023.01); G06N 3/08 (2023.01); G06Q 10/0631 (2023.01); G06Q 10/105 (2023.01); G06Q 30/0202 (2023.01); G06V 30/414 (2022.01); G06Q 50/20 (2012.01)
CPC G06Q 30/0202 (2013.01) [G06F 9/445 (2013.01); G06F 40/40 (2020.01); G06N 3/044 (2023.01); G06N 3/08 (2013.01); G06Q 10/06315 (2013.01); G06Q 10/105 (2013.01); G06V 30/414 (2022.01); G06Q 50/2057 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A method of improving the technological process of managing computer resources by reconfiguring an information technology (IT) system to meet future demand based on accurate inferences about said future demand, said method comprising:
obtaining, as input, in electronic form, structured information including tasks for a plurality of occupations in a plurality of industries related to information technology, over a length of time;
computing, from said structured information, a time series of normalized occupation task shares over said length of time, the computing including dividing counts of mentions of tasks by counts of mentions of occupations;
training a computerized machine learning model, on said time series, to predict future task shares for said plurality of occupations in said plurality of industries related to information technology, by applying an autoregressive integrated moving average (ARIMA) model wherein an architecture of the computerized machine learning model comprises an encoder that represents an intermediate layer implemented with a long short-term memory (LSTM) configured to transform the time series into another variable to capture couplings and parameters, and a decoder configured to produce the future task shares predictions based on an output of the encoder;
with said trained computerized machine learning model, predicting said future task shares and future demand for different types of computer-related tasks; and
reconfiguring said information technology system based on said predicted future demand for different types of computer-related tasks by taking at least one action from the group consisting of:
deploying additional hardware resources in a cloud computing environment; and
deploying additional software resources in a cloud computing environment.