Method for evaluating cleaning tasks based on action prediction model

Abstract

The embodiment of the application relates to a kind of based on action prediction model to the method for evaluating cleaning task, comprising: receiving the first data report sent by smart watch, contain first monitoring data, task start-stop time, seven kinds of long-time data sequence and the first task template customized for first cleaning personnel.Processing report extracts key information, obtains task place location from template, and sets task period according to task time.Seven kinds of long-time data sequence are processed into multi-modal data sequence, with task flow track node as subtask node, set subtask period and slice data.Subtask data sequence is input into action prediction model to predict action type, calculate subtask place location and indoor coordinate, obtain the distance of position related to task, and identify action type matching condition.Integrate distance and matching state to judge subtask evaluation result, and all pass to be eligible for task evaluation.Finally, relevant information is composed into subtask data set, and report constitutes first evaluation report and is saved, realizes scientific and accurate evaluation of cleaning task.

Images