Hfrxdjt

I've been trying to create a binary classification model that predicts wether there will be a train delay based on the train and time. Here is a link to the data

The issue I'm having is that my accuracy goes to 94.07 in the first 5 epochs. Meanwhile, my class prediction will always be 0 and never 1.

From what I understand, this is "Accuracy Paradox". A symptom of Class Imbalance. To combat this, I implemented Kfold.

kfold = StratifiedKFold(n_splits=10,shuffle=True)
cvs_scores = []

for train,test in kfold.split(X,Y):

 history = model.fit(X[train],Y[train],epochs=50,batch_size=15, shuffle = False, verbose = 1)
 scores = model.evaluate(X[test],Y[test],verbose=0)

 print("%s: %.2f%%" % (model.metrics_names[1],scores[1]*100))
 cvs_scores.append(scores[1] * 100)

print("%.2f%% (+/- %.2f%%)" % (numpy.mean(cvs_scores),numpy.std(cvs_scores)))

No luck. Still had the same issue as before.

Here is how I import my data:

raw_data = pd.read_csv('MTA_DELAY_DATA_DUMP - Sheet1.csv')

X = raw_data.iloc[1:-2,0:2].dropna().values
Y = raw_data.iloc[1:-2,2:3].dropna().astype(int).values

My Model:

model = Sequential()
model.add(Dense(32, kernel_initializer='uniform', activation='relu',input_dim =2))
model.add(Dense(16, kernel_initializer='uniform', activation='relu'))
model.add(Dense(8, kernel_initializer='uniform', activation='relu'))
model.add(Dense(1, kernel_initializer='uniform', activation='sigmoid'))

model.compile(loss='binary_crossentropy', optimizer=Adam(lr=0.001),metrics=['accuracy']) 
history = model.fit(X,Y,epochs=150,batch_size=15, shuffle = False, verbose = 1)

I tried assigning class weights to balance the data out. Even manually deleting 0's in the data, but nothing seems to result in accurate predictions. Am I doing something wrong in the model, or is this simply data that cannot be utilized by machine learning?