Leading indicators and monitor business KPIs and metrics for preemptive action

A computer-implemented method of data analytics, comprising:

• receiving training data including target time series data and candidate time series data, wherein the candidate time series data includes data corresponding to each of a plurality of candidate metrics;
• initializing, by a processor, a machine learning model comprising a recurrent neural network (RNN), a temporal attention head, and a candidate attention head;
• training, by the processor using the training data, the machine learning model to predict time series data by updating temporal attention weights of the temporal attention head corresponding to a plurality of rolling time windows and by updating candidate attention weights of the candidate attention head corresponding to the plurality of candidate metrics;
• computing, using the machine learning model, a sequence of hidden state vectors using the RNN;
• computing, using the machine learning model, a temporal attention output based on the sequence of hidden state vectors using the temporal attention head;
• computing, using the machine learning model, a variable importance output based on the temporal attention output using the candidate attention head;
• computing, using the machine learning model, a leading indicator metric for the target time series data based on the temporal attention weights and the candidate attention weights;
• collecting additional training data based on the leading indicator metric;
• and
• updating, by the processor using the additional training data, parameters of the machine learning model.