keras implemented, install tensorflow and keras (probably in a virtual...

keras implemented, install tensorflow and keras (probably in a virtual environment, you can read that from the page) and run classifier.py; commented out printing in arr_to_class function of pandas_stuff.py

classifier.py

+# LSTM with Dropout for sequence classification in the IMDB dataset
+import json
+import os.path
+import numpy as np
+import pandas_stuff as pd
+import keras.preprocessing.text as kpt
+from keras.models import Sequential, model_from_json
+from keras.layers import Dense
+from keras.layers import LSTM
+from keras.layers import Dropout
+from keras.layers.embeddings import Embedding
+from keras.preprocessing import sequence
+from keras.preprocessing.text import Tokenizer
+from sklearn.model_selection import train_test_split
+
+__path__ = "C:\\txtcls\\project.dir\\classifier.py"
+
+top_words = 5000
+max_length = 1000
+
+def make_dict(df, row, outfile='dict.json'):
+
+    if os.path.exists(outfile):
+        # read in our saved dictionary
+        with open(outfile, 'r') as f:
+            dictionary = json.load(f)
+        return dictionary
+    
+    x = df[row]
+    
+    tokenizer = Tokenizer(num_words=top_words) # dictionary size = 5000
+    tokenizer.fit_on_texts(x)
+    
+    with open(outfile, 'w') as f:
+        json.dump(tokenizer.word_index, f) # save dictionary of word indices
+
+    return tokenizer.word_index
+
+def text_to_indices(text, dictionary):
+    return [(dictionary[word] if dictionary[word] < top_words else 0) for word in kpt.text_to_word_sequence(text)]
+
+# Returns the vectorized job posting and expected output from the dataframe df.
+def get_from_frame(df, row, cols, elem):
+    x = df[row] # get the data as text
+    y = df[cols].values # get expected output
+
+    dictionary = make_dict(df,row)
+    # Convert raw text data into (dictionary-)indexed format
+    x = np.asarray(text_to_indices(x[elem],dictionary))
+    if len(x) > max_length: x = np.resize(x, max_length)    
+    x = np.pad(x, pad_width=(max_length - len(x), 0), mode='constant') # pad array to max length
+    
+    return (x, y[elem]) # return the (x,y) pair at index elem
+
+def ML_lstm(df, row, cols, model=None):
+    x = df[row] # get the data as text
+    y = df[cols].values
+
+    dictionary = make_dict(df,row)
+    # Convert raw text data into (dictionary-)indexed format
+    x = np.asarray([text_to_indices(text,dictionary) for text in x])
+    # split train/test
+    x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=.1, shuffle=True)
+    
+    # truncate/pad input sequences
+    x_train = sequence.pad_sequences(x_train, maxlen=max_length)
+    x_test = sequence.pad_sequences(x_test, maxlen=max_length)
+    
+    # build
+    vector_len = 32
+    if model == None:
+        model = Sequential()
+        model.add(Embedding(top_words, vector_len, input_length=max_length))
+        model.add(Dropout(0.2))
+        model.add(LSTM(100))
+        model.add(Dropout(0.2))
+        model.add(Dense(5, activation='softmax')) # was originally sigmoid activation, try this new stuff out
+        model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
+        print(model.summary())
+    model.fit(x_train, y_train, epochs=2, batch_size=64)
+    # Final evaluation of the model
+    scores = model.evaluate(x_test, y_test, verbose=0)
+    print("Accuracy: %.2f%%" % (scores[1]*100))
+
+    return model
+
+
+# fix random seed for reproducibility
+np.random.seed(7)
+model = None
+
+y_cols = ["hs_ged","associates","bachelors","masters","phd_prof"]
+df = pd.load_df("training.csv")
+
+if os.path.exists('model.json') and os.path.exists('model.h5'):
+    loading = input("Model found, load and test? y/n: ")
+    if loading in ['y','Y','yes']:
+        # load previously-trained model
+        with open('model.json', 'r') as json_file:
+            loaded_model_json = json_file.read()
+            model = model_from_json(loaded_model_json)
+        model.load_weights('model.h5')
+else:
+    model = ML_lstm(df, 'body', y_cols) # let's try it
+
+# now we can run some tests
+print("Model generated / loaded")
+while 1:
+    val = input("Input a number from 0~8000 or so: ")
+    if not val.isdigit():
+        print("non-integer value, exiting...")
+        break
+    val = int(val)
+    x,y = get_from_frame(df, "body", y_cols, val)
+    x = np.asarray([x])
+    pred = model.predict(x) # get model's prediction of input sample
+    print("Element #%d actual y value = " % val, y)
+    print("Element #%d predicted y value = " % val, pred)
+
+# save everything
+should_save = input("Save network? (y/n): ")
+if should_save in ['y','Y','yes']:
+    model_json = model.to_json()
+    with open('model.json', 'w') as json_file:
+        json_file.write(model_json)
+
+    model.save_weights('model.h5')
+    print('Keras network model saved.')

model.json

+{"class_name": "Sequential", "config": {"name": "sequential_1", "layers": [{"class_name": "Embedding", "config": {"name": "embedding_1", "trainable": true, "batch_input_shape": [null, 1000], "dtype": "float32", "input_dim": 5000, "output_dim": 32, "embeddings_initializer": {"class_name": "RandomUniform", "config": {"minval": -0.05, "maxval": 0.05, "seed": null}}, "embeddings_regularizer": null, "activity_regularizer": null, "embeddings_constraint": null, "mask_zero": false, "input_length": 1000}}, {"class_name": "Dropout", "config": {"name": "dropout_1", "trainable": true, "rate": 0.2, "noise_shape": null, "seed": null}}, {"class_name": "LSTM", "config": {"name": "lstm_1", "trainable": true, "return_sequences": false, "return_state": false, "go_backwards": false, "stateful": false, "unroll": false, "units": 100, "activation": "tanh", "recurrent_activation": "hard_sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "recurrent_initializer": {"class_name": "Orthogonal", "config": {"gain": 1.0, "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "unit_forget_bias": true, "kernel_regularizer": null, "recurrent_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "recurrent_constraint": null, "bias_constraint": null, "dropout": 0.0, "recurrent_dropout": 0.0, "implementation": 1}}, {"class_name": "Dropout", "config": {"name": "dropout_2", "trainable": true, "rate": 0.2, "noise_shape": null, "seed": null}}, {"class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "units": 5, "activation": "softmax", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}]}, "keras_version": "2.2.4", "backend": "tensorflow"}
\ No newline at end of file

pandas_stuff.py

@@ -33,7 +33,7 @@ def nd_arr_to_class(nd_arr):
         classes.append(num)
         num_classes.add(num)
         if num > 6:
-            print(row)
+            #print(row)
             num_bad += 1
     print("Num Classes", len(num_classes))
     print("Num Bad", num_bad)