def evaluateRandomForest():
print("\nEvaluating Random Forest")
regResults.append(["Results for Random Forest"])
for data in regDatasets:
#Import the Dataset and separate X and y
data_to_test = "regression/" + data + '.csv'
X_before, y_before = importDataset(data_to_test)
count = 0
avg_explained_variance_score = 0
avg_max_error = 0
avg_mae = 0
avg_mse = 0
avg_r2_score = 0
for train, test in kfold.split(X_before):
print("Test:", count+1, "for", data_to_test)
X_train, X_test = X_before.iloc[train], X_before.iloc[test]
y_train, y_true = y_before[train], y_before[test]
#feature scaling
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)
# run algorithm
from randomForest import randomForest
rfModel = randomForest(X_train, y_train, X_test, y_true, X_before)
predictions = rfModel.getPredictions()
# get metrics
avg_explained_variance_score += metrics.explained_variance_score(y_true, predictions)
avg_max_error += metrics.max_error(y_true, predictions)
avg_mae += metrics.mean_absolute_error(y_true, predictions)
avg_mse += metrics.mean_squared_error(y_true, predictions)
avg_r2_score += metrics.r2_score(y_true, predictions)
count += 1
avg_explained_variance_score = avg_explained_variance_score / count
avg_max_error = avg_max_error / count
avg_mae = avg_mae / count
avg_mse = avg_mse / count
avg_r2_score = avg_r2_score / count
regResults.append(['', data_to_test, float(avg_explained_variance_score), float(avg_max_error),
float(avg_mae), float(avg_mse), float(avg_r2_score)])
print("Random Forest evaluation results")
print("Average explained variance score:", avg_explained_variance_score)
print("Average mean absolute error:", avg_mae)
print("Average mean squared error:", avg_mse)
print("Average r2 score:", avg_r2_score)
def evaluateANN():
regResults.append(["Results for ANN"])
for data in regDatasets:
#Import the Dataset and separate X and y
data_to_test = "regression/" + data + '.csv'
X_before, y_before = importDataset(data_to_test)
count = 0
avg_explained_variance_score = 0
avg_max_error = 0
avg_mae = 0
avg_mse = 0
avg_r2_score = 0
for train, test in kfold.split(X_before):
print("Test:", count+1, " for", data_to_test)
X_train, X_test = X_before.iloc[train], X_before.iloc[test]
y_train, y_true = y_before[train], y_before[test]
#feature scaling
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)
# run ANN
from regressionAnalysis import sequentialNN
regressor = sequentialNN(X_train, y_train, X_test, y_true)
exp_variance_score, max_error, loss, mae, mse = regressor.getEvaluationMetrics()
# get metrics
avg_explained_variance_score += exp_variance_score
avg_max_error += max_error
avg_mae += mae
avg_mse += mse
avg_r2_score += metrics.r2_score(y_true, regressor.getPredictions())
count += 1
avg_explained_variance_score = avg_explained_variance_score / count
avg_max_error = avg_max_error / count
avg_mae = avg_mae / count
avg_mse = avg_mse / count
avg_r2_score = avg_r2_score / count
regResults.append(['', data_to_test, float(avg_explained_variance_score), float(avg_max_error),
float(avg_mae), float(avg_mse), float(avg_r2_score)])
print("ANN evaluation results")
print("Average explained variance score:", avg_explained_variance_score)
print("Average mean absolute error:", avg_mae)
print("Average mean squared error:", avg_mse)
print("Average r2 score:", avg_r2_score)
def selectDataset():
print("Select a file to use:")
print("1 - Regression Original 1437 rows")
print("2 - Regression Balanced (83% deleted)")
print("3 - Regression Encoded variables all 1437 rows")
print("4 - Regression Encoded variables balanced (83% deleted)")
print("5 - Regression no 365 days")
print("6 - Regression only 365 days")
print("7 - Regression only synthetic 3211 rows")
print("8 - Regression synthetic plus 365 days")
print("9 - Another dataset")
number = 0
acceptedDataset = False
while acceptedDataset is False:
number = int(input("Select number to import dataset: "))
if number > 0 and number < 10:
acceptedDataset = True
else:
print(
"Invalid number, select a dataset by selecting its number (1 to 9)"
)
choice = ""
if number == 1: choice = "regression/regAll.csv"
if number == 2: choice = "regression/regBalanced.csv"
if number == 3: choice = "regression/regEncoded.csv"
if number == 4: choice = "regression/regEncodedBalanced.csv"
if number == 5: choice = "regression/regNo365.csv"
if number == 6: choice = "regression/regOnly365.csv"
if number == 7: choice = "regression/regSynthetic.csv"
if number == 8: choice = "regression/regSyntheticWith365.csv"
if number == 9: choice = input("input full path of dataset: ")
print("dataset chosen:", choice)
# Import the Dataset and separate X and y
X_before, y_before = importDataset(choice)
# Split the dataset
X_train, X_test, y_train, y_test = splitAndScale(X_before, y_before)
return X_before, y_before, X_train, X_test, y_train, y_test, choice
def findBestMatch():
print("\nEvaluating different recipients")
X_before, y_before = importDataset('regression/regSyntheticWith365.csv')
X_train, X_test, y_train, y_true = splitAndScale(X_before, y_before)
# Train models with synthetic dataset
from regressionAnalysis import sequentialNN
sequentialNN(X_train, y_train, X_test, y_true)
ann = tf.keras.models.load_model('models/ann.h5')
from randomForest import randomForest
randomForest(X_train, y_train, X_test, y_true, X_before)
rf = joblib.load('models/rf.sav')
from svr import svr
svr(X_train, y_train, X_test, y_true)
svr = joblib.load('models/svr.sav')
MLmodels = [ann, rf, svr]
for data in recipientDatasets:
predict_results.append([data])
print("Predicting for",data)
dataset = pd.read_csv('datasets/' + data + '.csv')
to_predict = dataset.iloc[:, :-1].values
count = 1
for row in to_predict:
transform = scaler.fit_transform(row.reshape(-1, 1))
prediction = ['','donor'+ str(count)]
for model in MLmodels:
new_pred = model.predict(transform.reshape(1, -1))
if 'Sequential' in str(type(model)):
prediction.append(new_pred[0][0])
else:
prediction.append(new_pred[0])
predict_results.append(prediction)
count += 1
print('Predictions saved to file RecipientsPredictions.csv')