Method and system for predicting task completion of a time period based on task completion rates and data trend of prior time periods in view of attributes of tasks using machine learning models

Patent metadata

Field	Value
Publication number	US-2018150783-A1
Application number	US-201815882993-A
Country	US
Kind code	A1
Filing date	Jan 29, 2018
Priority date	Aug 24, 2016
Publication date	May 31, 2018
Grant date	—

Abstract

A request is received for determining a task completion rate of each of a first set of tasks associated with a set of task attributes. The first set of tasks are scheduled to be completed within a first timer period. An MAPE score is calculated or obtained for each of the completion rate predictive models, which is determined based on prior predictions performed in a second time period in the past. The duration of the second time period is a multiple of the first time period. One of the predictive models is selected based on the MAPE scores of the predictive models, where the selected model has the lowest MAPE score amongst the predictive models in the set. In another embodiment, a predictive model is selected further based on the volatility scores of the predictive models. A model with a combination of lowest MAPE score and volatility score is selected.

First claim

What is claimed is:

1 . A computer-implemented method for predicting task completion of a plurality of tasks, the method comprising:
receiving a request from a client for determining a task completion rate of each of a first set of a plurality of tasks associated with a plurality of attributes, the first set of tasks scheduled to be performed within a first time period; obtaining a mean absolute percentage error (MAPE) score for each of a set of predictive models based on predictions of each model in a second time period in the past, wherein the predictive models comprise one or more of a conversion-based probability model, task score aggregation model, a time series model, a neural network model, and a hybrid model; selecting one of the predictive models having a minimum MAPE score amongst the MAPE scores of the predictive models; and predicting a number of the first set of tasks to be completed in the first time period using the selected predictive model.

2 . The method of claim 1 , wherein the MAPE score of a predictive model is determined based on a difference between a predicted number of tasks to be completed during the second period of time and an actual number of tasks that actually have been completed within the second period of time.

3 . The method of claim 2 , wherein the difference between the predicted number of completed tasks and actual number of completed tasks is calculated for a plurality of third timer periods collectively constituting the second time period, and wherein the MAPE score is calculated based on the differences of the third time periods.

4 . The method of claim 3 , further comprising calculating a volatility score for each of the model based on the differences between the predicted and actual number of completed tasks between each pair of adjacent third time periods.

5 . The method of claim 1 , wherein the conversion-based probability model is to determine a probability of a set of tasks that have been initiated during the first period of time in view of a completion rate of tasks at the same time of the second time period.

6 . The method of claim 1 , wherein the task score aggregation model is to assign a task score for each task based on a plurality of attributes of the task, the task score representing a probability of the task that can be completed within the first time period.

7 . The method of claim 6 , wherein the task score is determined based on a stage of the task and a number of day remained until an end of the first time period, wherein each task goes through a plurality of stages before being completed.

8 . The method of claim 1 , wherein the neural network model is to predict a completion rate of tasks based on a plurality of attributes associated with the tasks including a number of days remained until an end of the first time period, a number of tasks in each of a plurality of stages, a number of new tasks created since starting of the first time period.

9 . The method of claim 1 , wherein the hybrid model is to predict a completion rate of a task using a weighted combination of a regression model and the conversion-based probability model or a weighted combination of the regression model the task score aggregation model.

10 . The method of claim 9 , wherein the regression model comprises one of a linear regression model, an exponential regression algorithm, and an autoregressive integrated moving average (ARIMA) algorithm.

11 . A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations of predicting task completion of a plurality of tasks, the operations comprising:
receiving a request from a client for determining a task completion rate of each of a first set of a plurality of tasks associated with a plurality of attributes, the first set of tasks scheduled to be performed within a first time period; obtaining a mean absolute percentage error (MAPE) score for each of a set of predictive models based on predictions of each model in a second time period in the past, wherein the predictive models comprise one or more of a conversion-based probability model, task score aggregation model, a time series model, a neural network model, and a hybrid model; selecting one of the predictive models having a minimum MAPE score amongst the MAPE scores of the predictive models; and predicting a number of the first set of tasks to be completed in the first time period using the selected predictive model.

12 . The machine-readable medium of claim 11 , wherein the MAPE score of a predictive model is determined based on a difference between a predicted number of tasks to be completed during the second period of time and an actual number of tasks that actually have been completed within the second period of time.

13 . The machine-readable medium of claim 12 , wherein the difference between the predicted number of completed tasks and actual number of completed tasks is calculated for a plurality of third timer periods collectively constituting the second time period, and wherein the MAPE score is calculated based on the differences of the third time periods.

14 . The machine-readable medium of claim 13 , wherein the operations further comprise calculating a volatility score for each of the model based on the differences between the predicted and actual number of completed tasks between each pair of adjacent third time periods.

15 . The machine-readable medium of claim 11 , wherein the conversion-based probability model is to determine a probability of a set of tasks that have been initiated during the first period of time in view of a completion rate of tasks at the same time of the second time period.

16 . The machine-readable medium of claim 11 , wherein the task score aggregation model is to assign a task score for each task based on a plurality of attributes of the task, the task score representing a probability of the task that can be completed within the first time period.

17 . The machine-readable medium of claim 16 , wherein the task score is determined based on a stage of the task and a number of day remained until an end of the first time period, wherein each task goes through a plurality of stages before being completed.

18 . The machine-readable medium of claim 11 , wherein the neural network model is to predict a completion rate of tasks based on a plurality of attributes associated with the tasks including a number of days remained until an end of the first time period, a number of tasks in each of a plurality of stages, a number of new tasks created since starting of the first time period.

19 . The machine-readable medium of claim 11 , wherein the hybrid model is to predict a completion rate of a task using a weighted combination of a regression model and the conversion-based probability model or a weighted combination of the regression model the task score aggregation model.

20 . The machine-readable medium of claim 19 , wherein the regression model comprises one of a linear regression model, an exponential regression algorithm, and an autoregressive integrated moving average (ARIMA) algorithm.

21 . A data processing system, comprising:
a processor; and a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations of predicting task completion of a plurality of tasks, the operations including
receiving a request from a client for determining a task co

Assignees

• Clari Inc

Inventors

• Xu Xin
• Tang Lei
• Rangan Venkat

Classifications

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/02

  Neural networks · CPC title