private static void MakePrediction(RFPClassifierLabel labelService, string Agency, string ClassCode, string Naics, string Subject, string SolNbr, string SetAside)
{
string prediction = labelService.PredictCategory(new RFPData
{
Agency = Agency,
ClassCode = ClassCode,
Naics = Naics,
Subject = Subject,
SolNbr = SolNbr,
SetAside = SetAside
});
Console.WriteLine($"Given Agency: {Agency} => Predicted: {prediction}");
}
public static void Main(string[] args)
{
/* Load directory containing training data
* iterate through all files and save model */
Console.WriteLine("Loading training data...");
TermCompiler TermsList = new TermCompiler(); // This instantiates class that can compile list of data for NER Tagging
string[] FileEntries = Directory.GetFiles("../../../Data"); // The Data folder in the solution directory
// These values will print as training occures to show how many files have been processed of total
int counter = 1;
int totalFiles = FileEntries.Length;
foreach (string Filename in FileEntries)
{
// Uncomment the following to train the model on the data in the Data directory
// TrainModel(Filename, counter, totalFiles);
// Will run through all the data and create a file with unique entries found in the data set (i.e. it will only write something once)
// 'TermCompiler' class can be modified to output whichever fields you want a list of (Agencies, Offices, Naics...)
TermsList.TermListCompiler(Filename);
counter++;
}
// This is an example of how you can test the model's accuracy with predictions
var labelService = new RFPClassifierLabel();
labelService.LoadModel(_modelPath);
Console.WriteLine("------------------------------------------------------");
Console.WriteLine("Predict some Offices based on other fields provided...");
Console.WriteLine("------------------------------------------------------");
Console.WriteLine();
// Should be --> DLA Acquisition Locations
MakePrediction(labelService, "Defense Logistics Agency", "15", "336413", "1560; PANEL STRUCTURAL, AI; T-38 ACFT; WSDC: 42(F); WSIC: T", "SPE4A719R0987", "N/A");
//I'm working on writing some code that can use built in algorithms + a unique testing dataset to show some accuracy metrics
}