Add: [AI-ML] Sentiment-Analyzer-Using-Neural-Networks

Domains/AI-ML/MiniProjects/SentimentAnalyzerUsingNeuralNetworks/dataset.csv

@@ -0,0 +1,41 @@
+review,sentiment
+"What a fantastic comeback by Liverpool!",positive
+"The referee ruined the match completely.",negative
+"It was an okay game, nothing special.",neutral
+"Messi’s goal was unbelievable!",positive
+"The defense was terrible today.",negative
+"The match ended in a draw, fair enough.",neutral
+"Ronaldo carried the whole team again!",positive
+"Our midfield looked clueless tonight.",negative
+"Good tactical game from both sides.",positive
+"Not the most exciting match to watch.",neutral
+"The keeper made some world-class saves!",positive
+"We wasted too many chances up front.",negative
+"Both teams deserved the point.",neutral
+"Brilliant performance by the youngsters!",positive
+"The coach has no idea what he's doing.",negative
+"The atmosphere in the stadium was electric!",positive
+"It was a boring goalless draw.",neutral
+"Terrible refereeing cost us the game.",negative
+"The fans were amazing throughout the match.",positive
+"Nothing went our way today.",negative
+"The new signing looked promising.",positive
+"The first half was dull but improved later.",neutral
+"We dominated possession but couldn’t score.",negative
+"Superb hat-trick from Haaland!",positive
+"The team lacked motivation and energy.",negative
+"The substitution made a huge impact.",positive
+"VAR decisions were so inconsistent again.",negative
+"Overall, a decent performance from both sides.",neutral
+"Best match of the season so far!",positive
+"That red card completely changed the game.",negative
+"Fair result, both teams played well.",neutral
+"Insane dribbling and pace from Mbappe!",positive
+"The defense collapsed after halftime.",negative
+"Not bad, but expected better from them.",neutral
+"Unbelievable last-minute winner!",positive
+"Sloppy passing all over the field.",negative
+"The crowd was quiet most of the time.",neutral
+"Fantastic teamwork and creativity today!",positive
+"Poor finishing cost us again.",negative
+"It was a balanced game overall.",neutral

Domains/AI-ML/MiniProjects/SentimentAnalyzerUsingNeuralNetworks/main.py

@@ -0,0 +1,63 @@
+import pandas as pd # for data manipulation
+from sklearn.model_selection import train_test_split # for splitting data
+# from sklearn.feature_extraction.text import CountVectorizer # for turning text into vectors
+# from sklearn.linear_model import LogisticRegression # for classification on data
+# from sklearn.metrics import accuracy_score # for evaluating model performance
+from sklearn.feature_extraction.text import TfidfVectorizer # checks how important a word is to a document in a collection 
+from sklearn.preprocessing import LabelEncoder # converts nuetral, positive and negative to values
+from tensorflow.keras.models import Sequential # for creating neural network model
+from tensorflow.keras.layers import Dense, Dropout # dense helps to build a neural network dropout enseures no overfitting
+
+# Load dataset
+data = pd.read_csv("dataset.csv")
+
+# train-test split:
+# splitting 80-20 for training and testing. this ensures that model would be able to genralize well on unseen data.
+# random state locks the data on randomizing everytime we run the code
+X_train, X_test,y_train, y_test = train_test_split(data['review'], data["sentiment"],test_size=0.2, random_state=42)
+
+# text vecotrization:
+# converts text data into numerical vectors(row, column format) that machine learning models can understand.
+# the tfidf was upgraded and max_featues keeps 5000 top most relevant words
+# vectorizer = CountVectorizer()
+vectorizer = TfidfVectorizer(max_features=5000)
+
+# fit transform learns vocabulary while transform uses same vocabulary data
+X_train_vectors = vectorizer.fit_transform(X_train)
+X_test_vectors = vectorizer.transform(X_test)
+
+# Update encode labels
+encoder = LabelEncoder()
+y_train_enc = encoder.fit_transform(y_train)
+y_test_enc = encoder.transform(y_test)
+
+# Model training:
+# model = LogisticRegression()
+# # learning from training data X_train_vectors the vectored sentence and y_train the labels
+# model.fit(X_train_vectors, y_train)
+# upgrade building a neural network
+model = Sequential([
+    Dense(128, activation='relu', input_shape=(X_train_vectors.shape[1],)),
+    Dropout(0.3),
+    Dense(64, activation='relu'),
+    Dense(3, activation='softmax')  # 3 classes: positive, negative, neutral
+])
+
+#compile the model
+model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
+
+#Train model
+model.fit(X_train_vectors.toarray(),y_train_enc, epochs=10, batch_size=32, validation_split=0.2)
+
+# Evaluation
+# predictions = model.predict(X_test_vectors)
+# accuracy = accuracy_score(y_test, predictions)
+loss, accuracy = model.evaluate(X_test_vectors.toarray(), y_test_enc)
+print(f"Model Accuracy: {accuracy*100:.2f}%")
+
+# Prediction:
+user_input = input("Enter a football match review: ")
+user_vector = vectorizer.transform([user_input]).toarray()
+prediction = model.predict(user_vector)
+sentiment = encoder.inverse_transform([prediction.argmax()])[0]
+print(f"Predicted Sentiment: {sentiment.capitalize()}")