System and method for best-practice-based budgeting

1. A computer-implemented method for determining a status of an intellectual property case, the computer-implemented method comprising:

obtaining, over a computer network from a plurality of customers, a plurality of file history action (FHA) designations for a plurality of cases manually specified by employees of the customers;

identifying, over the computer network for each of the cases:

a date for a most recent case event, and

a set of document codes for the date, the document codes corresponding to documents associated with the date and indicating what corresponding documents represent;

for each obtained FHA designation, forming a plurality of document code bit vectors based on the document codes identified for cases with the obtained FHA designation;

training a model that receives, as input, a document code bit vector of document codes, and that produces, as output, a predicted FHA designation and a confidence score corresponding to the predicted FHA designation, the training comprising employing at least one of a hidden Markov model and a multiclass support vector machine;

for each case in a portfolio of cases:

determining, at least in part by performing screen-scraping of a website of a remote data provider, that the case has a case event;

identifying, from the remote data provider, a plurality of document codes for a date of the case event;

forming a document code bit vector of the document codes identified for the date of the case event;

providing the document code bit vector as input to the trained model;

obtaining a predicted FHA designation and a confidence score from the trained model;

determining that the confidence score is above a given threshold; and

responsive to the confidence score being above the given a threshold, predicting a remaining cost to completion of the case using the predicted FHA designation; and

using the predicted costs to completion of the cases to predict a total cost to completion of the portfolio.

2. The computer-implemented method of claim 1 , wherein determining, over the computer network, that a case has a case event comprises receiving an email.

3. The computer-implemented method of claim 1 , further comprising identifying additional possible documents codes by analyzing document codes in web pages of the remote data provider.

4. The computer-implemented method of claim 3 , further comprising retraining the model in response to identifying the additional possible document codes.

5. The computer-implemented method of claim 1 , further comprising computing, based on a plurality of completed cases, statistical averages for costs to completion from an event corresponding to a predicted FHA designation.

6. The computer-implemented method of claim 5 , further comprising predicting the remaining costs to completion of the cases using the computed statistical averages.

7. A computer-implemented method comprising:

obtaining a plurality of file history action (FHA) designations for a plurality of cases;

identifying a set of document codes corresponding to the obtained FHA designations;

for each obtained FHA designation, forming a plurality of document code vectors based on the document codes identified for cases with the obtained FHA designation;

training a model that receives, as input, a document code vector of document codes, and that produces, as output, an estimated status and a confidence score corresponding to the estimated status designation, the training comprising employing at least one of a hidden Markov model and a multiclass support vector machine;

for each portfolio case in a portfolio:

determining, at least in part by performing screen-scraping of a website of a remote data provider, that the portfolio case has a case event;

identifying a plurality of document codes for a date of the case event, the document codes corresponding to documents associated with the date and indicating what corresponding documents represent;

forming a document code bit vector of the document codes identified for the date of the case event;

providing the document code bit vector as input to the trained model;

obtaining a first predicted FHA designation from the trained model; and

using the first predicted FHA designation, predicting a remaining cost to completion of the portfolio case; and

using the predicted remaining costs to completion of the portfolio cases to predict a total cost to completion of the portfolio.

8. The computer-implemented method of claim 7 , further comprising computing, based on a plurality of completed cases, statistical averages for costs to completion from an event corresponding to the first predicted FHA designation.

9. The computer-implemented method of claim 8 , further comprising predicting the remaining cost to completion of the portfolio cases using the computed statistical averages.

10. The computer-implemented method of claim 7 , further comprising identifying additional possible documents codes by analyzing document codes in web pages of a remote data provider from which the set of documents codes was identified.

11. The computer-implemented method of claim 10 , further comprising retraining the model in response to identifying the additional possible documents codes.

12. The computer-implemented method of claim 7 , wherein determining, over a computer network, that a case has a case event comprises receiving an email.

13. The computer-implemented method of claim 7 , further comprising:

determining that the first confidence score is above a given a threshold; and

responsive to the first confidence score being above a given a threshold:

identifying a docket item associated with the first predicted FHA designation,

scheduling the first docket item in association with the first case, and

notifying an owner of the first case of the first docket item.