Machine learning classifier for identifying internet service providers from website tracking

1. A method being performed by a computer system that comprises one or more processors and a computer-readable storage medium encoded with instructions executable by at least one of the processors and operatively coupled to at least one of the processors, the method comprising:

obtaining website data traffic content for a website including a plurality of IP addresses for visitors visiting the website;

mapping the IP addresses to a plurality of business entity identifiers to identify a plurality of business entities associated with the website data traffic data;

obtaining attribute data for each business entity associated with each of the plurality of business entity identifiers to identify business attributes associated with each of the business entities mapped to the each of the IP addresses;

generating a training data set from the website data traffic content and the attribute data for training an Internet Service Provider (ISP) classifier to identify ISPs associated with the mapped website traffic data, the training data set being generated from website data traffic content having IP addresses successfully matched to the business entities identifiers, and the attribute data comprises a feature set for a model comprising features from firmographic data for the attribute data and the web traffic data comprising: industry code, unique visitors, visit count, employee count, revenue, and a customer data;

wherein the web traffic data comprises, per IP address, a count of visits, unique visitors, and index customer count that accessed the website within a time window;

wherein the time window comprise an interval of: ingesting the web traffic data,

mapping the web traffic data,

running the classifier to identify ISP and non-ISP web traffic; and

configuring the continuous maintenance of the model by aggregate tracking and machine intelligence;

generating a model for training the ISP classifier,

wherein the model comprises a supervised classification,

wherein the supervised classification model is a gradient boosting classifier; and

wherein the gradient boosting classifier training comprises iterating the gradient booster for about 1000 iterations,

training the model on the training data set;

determining if the trained model accurately identifies ISPs; and

if the model is accurate, applying the trained ISP classifier to a database of website content data; and

identifying website traffic for the website attributable to ISPs.

2. The method of claim 1 , further comprising:

obtaining a match acceptance rate of IP addresses from the website traffic data mapped to a plurality of the business entities identifiers;

generating the training data set from the IP addresses from the website traffic data matched to the business entities identifiers.

3. The method of claim 2 , further comprising:

replacing the IP addresses from the website traffic data mapped to the business entities identifiers with previously identified IP addresses from the website traffic data that were previously mapped or matched to other business entities.

4. The method of claim 1 , further comprising:

generating the model for training the ISP classifier; and

if the model is accurate, applying the trained model as the ISP classifier to the database of website content data.

5. The method of claim 4 , further comprising:

applying the trained ISP classifier to a database of IP addresses mapped to business entity identifications from a business entity firmographic database that does not include website traffic visit information and website traffic visitor information.

6. The method of claim 1 , wherein obtaining website data traffic content including IP addresses data for a website further comprises:

placing a tracking code object on a website to be tracked;

generating a log file identifying all IPs and cookies associated with the IPs tracked by the tracking code.

7. The method of claim 1 , wherein the time window comprises:

a time window selected from the group consisting of: a weekly time interval, a biweekly interval, a monthly interval, a bi-monthly interval, and a quarterly interval.

8. The method of claim 7 , wherein the time window comprises: the monthly interval.

9. The method of claim 1 , wherein the revenue data comprises annual sales.

10. The method of claim 1 , wherein the industry code comprises an SIC code, an NAICS code, or both.

11. The method of claim 1 , wherein the gradient boosting classifier model {F} that predict values of the form y=F(x) by minimizes g the mean squared error.

12. The method of claim 1 , wherein the gradient boosting classifier comprises:

at each stage m of gradient boosting (1<=m<=M) for model Fm predicting mean y in the training set, the gradient boosting algorithm adds an estimator h [Fm+1(x)=Fm(x)+h(x)], where Fm+1(x)=Fm(x)+h(x)=y, wherein the gradient boosting fits h to the residual y−Fm(x).

13. A computer program stored on a computer readable medium and loadable into the internal memory of a digital computer, comprising software code portions, when said program is run on a computer, for performing the method of claim 1 .

14. The method of claim 1 , wherein the gradient boosting classifier model substantially fits the data.

15. The method of claim 14 , wherein the gradient boosting classifier model fits the data and produces substantially accurate ISP identification than other classifier models.

16. The method of claim 1 , wherein the gradient boosting classifier further comprises processing on real web traffic data and results impact directly, while employing automated notification gates at stages.

17. The method of claim 16 , wherein the gradient boosting classifier is configured to provide automated notifications if the confusion matrix or output results are not accurate.

18. The method of claim 1 , wherein the gradient boosting classifier further comprises a monthly interval run of the classifier producing output classification output.

19. The method of claim 18 , wherein the gradient boosting classifier further comprises a manual review and consolidation of records.

20. The method of claim 19 , wherein the gradient boosting classifier further comprises a revised dataset, appended to a next monthly interval cycle of the model run and update the training of the model using machine intelligence.

21. A computer system comprising:

a network computer, including:

a transceiver for communicating over the network;

a memory for storing at least instructions and a word database; and

a processor device that is operative to execute instructions that enable actions, comprising:

obtaining website data traffic content including IP address data for a website;

mapping the IP address data for the website to at least one business entity identifier to identify at least one business entity associated with the website data traffic data;

obtaining attribute data for each business entity associated with the at least one business entity identifier to identify business attributes associated with the at least one business entity mapped to the IP address data;

generating a training data set from the website data traffic content and the attribute data for training an Internet Service Provider (ISP) classifier to identify ISPs associated with the mapped website traffic data, the training data set being generated from website data traffic content having IP addresses successfully matched to the business entities identifiers, and the attribute data comprises a feature set for a model comprising features from firmographic data for the attribute data and the web traffic data comprising: industry code, unique visitors, visit count, employee count, revenue, and a customer data;

wherein the web traffic data comprises, per IP address, a count of visits, unique visitors, and index customer count that accessed the website within a time window;

wherein the time window comprise an interval of: ingesting the web traffic data,

mapping the web traffic data,

running the classifier to identify ISP and non-ISP web traffic; and

configuring the continuous maintenance of the model by aggregate tracking and machine intelligence;

generating a model for training the ISP classifier,

wherein the model comprises a supervised classification;

wherein the supervised classification model is a gradient boosting classifier; and

wherein the gradient boosting classifier training comprises iterating the gradient booster for about 1000 iterations,

training the model on the training data set;

determining if the trained model accurately identifies ISPs; and

if the model is accurate, applying the trained ISP classifier to a database of website content data; and

identifying website traffic for the website attributable to ISPs.