Multi-step day sales outstanding forecasting

One or more non-transitory computer-readable storage media coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to:

• receive historical financial line item data;
• extract a historical time series from the historical financial line item data;
• generate and train a signal processing model using the historical time series;
• receive open receivables financial line item data for transactions that posted before an end of a selected time period for converting open receivables to revenue and that remained open after the selected time period;
• receive revenue financial line item data;
• provide the open receivables financial line item data to a day sales outstanding (DSO) predictor engine to generate a predicted open receivables comprising a multi-step time series forecasting regression generated from the open receivables financial line item data, wherein the DSO predictor engine performs operations comprising:
• receiving financial line item data;
• extracting item features from the financial line item data;
• generating a signal processed time series by applying the trained signal processing model to identify patterns within the financial line item data and transform the financial line item data in a lower-dimensional space;
• clustering each of the signal processed time series to an optimal non-overlapping cluster;
• training a regression model for each time point in each cluster based on the time series in the cluster;
• and
• generating the multi-step time series forecasting regression by applying a respective trained regression model to each future time point in each of the optimal non-overlapping clusters to predict a time series value for each future time point;
• provide the revenue financial line item data to the DSO predictor engine to generate a predicted revenue comprising the multi-step time series forecasting regression generated from the revenue financial line item data;
• generate a predicted DSO based on the predicted open receivables and the predicted revenue, wherein the predicted DSO includes a vector of final DSO predictions for each future time point;
• determine that the predicted DSO is greater than a threshold;
• and
• increase the selected time period for converting open receivables to revenue based on determining that the predicted DSO is greater than the threshold.

A day sales outstanding (DSO) forecasting system, comprising:

• one or more processors;
• and
• a computer-readable storage device coupled to the one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to:
• receive historical financial line item data;
• extract a historical time series from the historical financial line item data;
• generate and train a signal processing model using the historical time series;
• receive open receivables financial line item data for transactions that posted before an end of a selected time period for converting open receivables to revenue and that remained open after the selected time period;
• receive revenue financial line item data;
• provide the open receivables financial line item data to a DSO predictor engine to generate a predicted open receivables comprising a multi-step time series forecasting regression generated from the open receivables financial line item data, wherein the DSO predictor engine performs operations comprising:
• receiving revenue financial line item data;
• extracting item features from the financial line item data;
• generating a signal processed time series by applying the trained signal processing model to identify patterns within the financial line item data and transform the financial line item data in a lower-dimensional space;
• clustering each of the signal processed time series to an optimal non-overlapping cluster;
• training a regression model for each time point in each cluster based on the time series in the cluster;
• and
• generating the multi-step time series forecasting regression by applying a respective trained regression model to each future time point in each of the optimal non-overlapping clusters to predict a time series value for each future time point;
• provide the revenue financial line item data to the DSO predictor engine to generate a predicted revenue comprising the multi-step time series forecasting regression generated from the revenue financial line item data;
• generate a predicted DSO based on the predicted open receivables and the predicted revenue, wherein the predicted DSO includes a vector of final DSO predictions for each future time point;
• determine that the predicted DSO is greater than a threshold;
• and
• increase the selected time period for converting open receivables to revenue based on determining that the predicted DSO is greater than the threshold.