private void btn_predict_Click(object sender, EventArgs e)
{
float sqFeet = 0;
if (!float.TryParse(txt_sqfeet.Text, out sqFeet) || sqFeet < 1)
{
MessageBox.Show("Please enter square feet value");
return;
}
float bedrooms = 0;
if (!float.TryParse(txt_bedrooms.Text, out bedrooms) || bedrooms < 1)
{
MessageBox.Show("Please enter bedroom count");
return;
}
float year = 0;
if (!float.TryParse(txt_YearBuilt.Text, out year) || year < 1)
{
MessageBox.Show("Please enter year built");
return;
}
float[] testData = { sqFeet, bedrooms, year, (float)Math.Sqrt(sqFeet), (float)Math.Sqrt(bedrooms), (float)Math.Sqrt(year), sqFeet *bedrooms };
uint featureCount = 7;
uint testDataCount = 1;
float[] predictionResult = new float[testDataCount];
//Call the GPU module
LR_GPULibProxy.Predict(testData, featureCount, testDataCount, hypothesis, mean, std, predictionResult);
labelPrice.Text = predictionResult[0].ToString("N0");
}
private void btn_trainLinearRegression_Click(object sender, EventArgs e)
{
//get the data
IList <HousingDTO> housingDataList = dgv_trainingData.DataSource as List <HousingDTO>;
if (dgv_trainingData.RowCount < 1 || housingDataList == null)
{
MessageBox.Show("Please load data");
return;
}
//convert data to array
int featureCount = 7;
float[] featureData = new float[housingDataList.Count * featureCount];
float[] labelData = new float[housingDataList.Count];
int idx = 0;
int labelIdx = 0;
//Add the polynomial terms
foreach (HousingDTO dto in housingDataList)
{
featureData[idx++] = dto.SquareFeet;
featureData[idx++] = dto.BedroomCount;
featureData[idx++] = dto.YearBuilt;
featureData[idx++] = (float)Math.Sqrt(dto.SquareFeet); //add polynomial term for better fit
featureData[idx++] = (float)Math.Sqrt(dto.BedroomCount); //add polynomial term for better fit
featureData[idx++] = (float)Math.Sqrt(dto.YearBuilt); //add polynomial term for better fit
featureData[idx++] = dto.SquareFeet * dto.BedroomCount;
labelData[labelIdx++] = dto.Price;
}
int trainingDataCount = labelData.Length;
float learningRate = float.Parse(txt_learningRate.Text);
uint iterations = uint.Parse(txt_Iterations.Text);
float regularizationTerm = float.Parse(txt_RegularizationTerm.Text);
//featureCount == 7 -> == 8
hypothesis = new float[featureCount + 1]; //add one for the bias term
mean = new float[featureCount + 1];
std = new float[featureCount + 1];
//Call the GPU module
LR_GPULibProxy.Learn(featureData, labelData, (uint)featureCount, (uint)trainingDataCount, iterations, learningRate, regularizationTerm, hypothesis, mean, std);
//Display the model
StringBuilder model = new StringBuilder();
model.Append(string.Format("{0:N2}", hypothesis[0]));
//featureCount == 7
for (int i = 1; i < featureCount + 1; i++)
{
model.Append(string.Format("{0}{1:N2}*X{2}", hypothesis[i] < 0 ? "" : "+", hypothesis[i], i));
}
labelHypothesis.Text = model.ToString();
btn_predict.Enabled = true;
}
