model 50 epochs

.gitignore

@@ -7,4 +7,5 @@ data_train.csv
 data.csv 
 data_not_shuf.csv
 data_not_cutted.csv
-venv
+venv
+.~lock.fake_job_postings.csv#

main.py

@@ -10,20 +10,29 @@ from torch import nn
 from torch import optim
 import matplotlib.pyplot as plt
 
+
+def convert_text_to_model_form(text):
+    a = vectorizer.transform([text])
+    b = torch.tensor(scipy.sparse.csr_matrix.todense(a)).float()
+    return b
+
+
 if __name__ == "__main__":
-    # kaggle.api.authenticate()
-    # kaggle.api.dataset_download_files('shivamb/real-or-fake-fake-jobposting-prediction', path='.',
-    #                                   unzip=True)
+    kaggle.api.authenticate()
+    kaggle.api.dataset_download_files('shivamb/real-or-fake-fake-jobposting-prediction', path='.',
+                                      unzip=True)
 
     data = pd.read_csv('fake_job_postings.csv', engine='python')
-    data = data.replace(np.nan, '', regex=True)
+    # data = data.replace(np.nan, '', regex=True)
+    data = data[["company_profile", "fraudulent"]]
+    data = data.dropna()
 
     data_train, data_test = train_test_split(data, test_size=3000, random_state=1)
     data_dev, data_test = train_test_split(data_test, test_size=1500, random_state=1)
 
-    x_train = data_train["title"]
-    x_dev = data_dev["title"]
-    x_test = data_test["title"]
+    x_train = data_train["company_profile"]
+    x_dev = data_dev["company_profile"]
+    x_test = data_test["company_profile"]
 
     y_train = data_train["fraudulent"]
     y_dev = data_dev["fraudulent"]
@@ -31,27 +40,32 @@ if __name__ == "__main__":
 
     x_train = np.array(x_train)
     x_dev = np.array(x_dev)
+    x_test = np.array(x_test)
 
     y_train = np.array(y_train)
     y_dev = np.array(y_dev)
+    y_test = np.array(y_test)
 
     vectorizer = TfidfVectorizer()
 
     x_train = vectorizer.fit_transform(x_train)
     x_dev = vectorizer.transform(x_dev)
+    x_test = vectorizer.transform(x_test)
 
     x_train = torch.tensor(scipy.sparse.csr_matrix.todense(x_train)).float()
     x_dev = torch.tensor(scipy.sparse.csr_matrix.todense(x_dev)).float()
+    x_test = torch.tensor(scipy.sparse.csr_matrix.todense(x_test)).float()
 
     y_train = torch.tensor(y_train)
     y_dev = torch.tensor(y_dev)
+    y_test = torch.tensor(y_test)
 
     from torch import nn
 
     model = nn.Sequential(
         nn.Linear(x_train.shape[1], 64),
         nn.ReLU(),
-        nn.Linear(64, data_train["title"].nunique()),
+        nn.Linear(64, data_train["fraudulent"].nunique()),
         nn.LogSoftmax(dim=1))
 
     # Define the loss
@@ -65,7 +79,7 @@ if __name__ == "__main__":
     test_losses = []
     test_accuracies = []
 
-    epochs = 5
+    epochs = 50
     for e in range(epochs):
         optimizer.zero_grad()
 
@@ -97,6 +111,50 @@ if __name__ == "__main__":
               f"Test Loss: {test_loss:.3f}.. ",
               f"Test Accuracy: {test_accuracy:.3f}")
 
+    TP = []
+    TF = []
+
+    FP = []
+    FN = []
+    log_ps = model(x_test)
+    ps = torch.exp(log_ps)
+    top_p, top_class = ps.topk(1, dim=1)
+    descr = np.array(data_test["company_profile"])
+    for i, (x, y) in enumerate(zip(np.array(top_class), np.array(y_test.view(*top_class.shape)))):
+        d = descr[i]
+        if x == y:
+            if x:
+                TP.append(d)
+            else:
+                TF.append(d)
+        else:
+            if x:
+                FP.append(d)
+            else:
+                FN.append(d)
+    f_score = len(TP) / (len(TP) + 0.5 * (len(FP) + len(FN)))
+    f = open("model_resutls.txt", "a")
+
+    f.write(f"F-SCORE = {f_score}\n")
+    f.write(f"TP = {len(TP)}\n")
+    f.write(f"TF = {len(TF)}\n")
+    f.write(f"FP = {len(FP)}\n")
+    f.write(f"FN = {len(FN)}\n")
+
+    f.write(f"TP descriptions:")
+    for i in TP:
+        f.write(i+'\n')
+    f.write(f"TF descriptions:")
+    for i in TF:
+        f.write(i+"\n")
+    f.write(f"FP descriptions:")
+    for i in FP:
+        f.write(i+"\n")
+    f.write(f"FN descriptions:")
+    for i in FN:
+        f.write(i+"\n")
+    f.close()
+
     plt.figure(figsize=(12, 5))
     ax = plt.subplot(121)
     plt.xlabel('epochs')
@@ -109,5 +167,3 @@ if __name__ == "__main__":
     plt.ylabel('test accuracy')
     plt.plot(test_accuracies)
     plt.show()
-
-    print('Succes')