Sentiment Analysis on Twitter Tweets

---

Sentiment Analysis is conducted on various datasets after exploratory data analysis and data preprocessing, separately using variety of Machine Learning techniques

15 implemented ML Algorithms :

1. Logistic Regression
• Newton CG
• SAG
• SAGA
• LBFGS
2. Decision Tree Classifier
3. Support Vector Machines
• Linear
• Poly
• RBF
• Sigmoid
4. Majority Voting Ensemble
5. Extreme Laerning Machines
• Tanh
• SinSQ
• Tribas
• Hardlim
6. Artificial Neural Networks (Multi - Layer Perceptron) Gradient Descent

---

Part 1 : Data Preprocessing

1. Exploratory Data Analysis
2. Data Preprocessing
3. Cleaning
4. Lemmatization

>>> Run Text_Preprocessing_MP_Hybrid.py if you want to Preprocess some datasets

---

Flow of Control

1. Sentence Segmentation
2. Word Tokenization
3. Same consecutive chars changed to max 2 times
4. Spelling Corrections
5. Removal of #Hashtags, @Mentions, http//:URLs, etc (Noise 1)
6. Removal of Special Unicode Characters (Noise 2)
7. Chat Abbreviations conversions (Noise 3)
8. Removal of Punctuations except `'` (Noise 4)
9. Stop Words Removal (Noise 5)
10. Parts of Speech Tagging
11. Stemming & Lemmatization
12. WhiteSpace Removals
13. Chunking

---

Part 2 : Machine Learning Models Training

---

Datasets

Dataset 1

Value Counts

• -1 : 15236
• 0 : 9465
• 1 : 15299

Dataset 2

Value Counts

• -1 : 1117
• 0 : 1570
• 1 : 1186

Dataset 3

Value Counts

• 0 : 29720
• 1 : 2242

Dataset 4

Value Counts

• -1 : 35510
• 0 : 55213
• 1 : 72250