chore: add links for codes, Jupyter notebooks and datasets in README.

Codes/Code-1.md

@@ -0,0 +1,186 @@
+# Practical-1 (Linear Regression using Deep Neural Network)
+
+Problem Statement: Linear regression by using Deep Neural network: Implement Boston housing price prediction problem by Linear regression using Deep Neural network. Use Boston House price prediction dataset.
+
+> [!NOTE]
+> Dataset available in [Datasets](../Datasets/boston.csv) directory.
+
+---
+
+## Pre-requisities
+
+1. Install packages using `pip`: `pip install tensorflow keras pandas numpy scikit-learn matplotlib seaborn` (`tensorflow` requires Python 3.9 - 3.12)
+2. Copy the `boston.csv` dataset in the same directory as the Jupyter notebook.
+
+## Steps
+
+1. Import Libraries
+2. Load Dataset
+3. Exploratory Data Analysis (EDA)
+4. Check for Missing Values
+5. Correlation Heatmap
+6. Separate Features and Target
+7. Split into Training and Testing Sets
+8. Feature Scaling (Standardization)
+9. Build the Neural Network Model
+10. Compile the Model
+11. Train the Model
+12. Evaluate the Model on Test Data
+13. Make Predictions
+14. Plot Training vs Validation Loss
+15. Plot Predicted vs Actual Prices
+
+---
+
+## Code
+
+### 1. Import Libraries:
+
+```python3
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import StandardScaler
+from keras import Input
+from keras.models import Sequential
+from keras.layers import Dense
+```
+
+### 2. Load Dataset:
+
+```python3
+data = pd.read_csv('boston.csv')
+print(data.head())
+```
+
+### 3. Exploratory Data Analysis (EDA):
+
+```python3
+print("Shape:", data.shape)          # number of rows and columns
+print("\nData Types:\n", data.dtypes)
+print("\nStatistical Summary:\n", data.describe())  # min, max, mean, std, etc.
+```
+
+### 4. Check for Missing Values:
+
+```python3
+print("Missing values per column:\n", data.isnull().sum())
+
+# Drop rows with missing values (if any)
+data = data.dropna()
+print("\nShape after dropping nulls:", data.shape)
+```
+
+### 5. Correlation Heatmap:
+
+```python3
+plt.figure(figsize=(12, 8))
+sns.heatmap(data.corr(), annot=True, fmt=".2f", cmap="coolwarm")  # show correlation between all feature pairs
+plt.title("Feature Correlation Heatmap")
+plt.tight_layout()
+plt.show()
+```
+
+### 6. Separate Features and Target:
+
+```python3
+X = data.drop('MEDV', axis=1)   # all columns except house price
+y = data['MEDV']                 # target: median house price
+```
+
+### 7. Split into Training and Testing Sets:
+
+```python3
+# 80% train, 20% test; random_state=42 ensures reproducible split
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
+```
+
+### 8. Feature Scaling (Standardization):
+
+```python3
+scaler = StandardScaler()
+
+X_train = scaler.fit_transform(X_train)  # learn mean/std from train, then scale
+X_test = scaler.transform(X_test)        # apply same mean/std to test (no leakage)
+```
+
+### 9. Build the Neural Network Model:
+
+```python3
+model = Sequential()
+model.add(Input(shape=(X_train.shape[1],)))  # input shape = number of features
+model.add(Dense(64, activation='relu'))       # hidden layer 1: 64 neurons
+model.add(Dense(32, activation='relu'))       # hidden layer 2: 32 neurons
+model.add(Dense(1, activation='linear'))      # output layer: single value (house price)
+
+model.summary()
+```
+
+### 10. Compile the Model:
+
+```python3
+# adam: adaptive optimizer; mse: standard regression loss; mae: human-readable error metric
+model.compile(optimizer='adam', loss='mean_squared_error', metrics=['mae'])
+```
+
+### 11. Train the Model:
+
+```python3
+# validation_split=0.2 reserves 20% of training data to monitor val loss each epoch
+history = model.fit(X_train, y_train, epochs=100, batch_size=32, validation_split=0.2)
+```
+
+### 12. Evaluate the Model on Test Data:
+
+```python3
+loss, mae = model.evaluate(X_test, y_test)
+print(f"Test Loss (MSE): {loss:.4f}")
+print(f"Test Mean Absolute Error: {mae:.4f}")
+```
+
+### 13. Make Predictions:
+
+```python3
+predictions = model.predict(X_test)
+print("First 5 Predicted Prices:", predictions[:5].flatten())
+print("First 5 Actual Prices:   ", y_test.values[:5])
+```
+
+### 14. Plot Training vs Validation Loss:
+
+```python3
+plt.plot(history.history['loss'], label='Training Loss')
+plt.plot(history.history['val_loss'], label='Validation Loss')
+plt.title('Model Loss Over Epochs')
+plt.ylabel('Loss (MSE)')
+plt.xlabel('Epoch')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 15. Plot Predicted vs Actual Prices:
+
+```python3
+plt.figure(figsize=(8, 6))
+plt.scatter(y_test, predictions, alpha=0.7)                        # each point = one test sample
+plt.plot([y_test.min(), y_test.max()],
+         [y_test.min(), y_test.max()], 'r--', label='Ideal Fit')   # diagonal = perfect prediction
+plt.xlabel('Actual Price')
+plt.ylabel('Predicted Price')
+plt.title('Actual vs Predicted House Prices')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+---
+
+## Miscellaneous
+
+- [Dataset source](https://www.kaggle.com/datasets/fedesoriano/the-boston-houseprice-data)
+
+---
+

Codes/Code-2b.md

@@ -0,0 +1,202 @@
+# Practical-2b (Classification using Deep Neural Network - IMDB Dataset)
+
+Problem Statement: Binary classification using Deep Neural Networks Example: Classify movie reviews into positive" reviews and "negative" reviews, just based on the text content of the reviews. Use IMDB dataset
+
+> [!NOTE]
+> Dataset available in [Datasets](../Datasets/IMDB%20Dataset.csv) directory.
+
+---
+
+## Pre-requisities
+
+1. Install packages using `pip`: `pip install tensorflow keras pandas numpy scikit-learn matplotlib seaborn` (`tensorflow` requires Python 3.9 - 3.12)
+2. Copy the `IMDB Dataset.csv` dataset in the same directory as the Jupyter notebook.
+
+## Steps
+
+1. Import Libraries
+2. Load Dataset
+3. Exploratory Data Analysis (EDA)
+4. Data Cleaning - Strip HTML Tags
+5. Encode Labels and Separate Features
+6. Tokenize and Pad Text Sequences
+7. Split into Training and Testing Sets
+8. Build the Neural Network Model
+9. Compile the Model
+10. Train the Model
+11. Evaluate the Model on Test Data
+12. Plot Training vs Validation Accuracy
+13. Plot Training vs Validation Loss
+14. Confusion Matrix and Classification Report
+
+---
+
+## Code
+
+### 1. Import Libraries:
+
+```python3
+import re
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+from sklearn.model_selection import train_test_split
+from sklearn.preprocessing import LabelEncoder
+from sklearn.metrics import confusion_matrix, classification_report
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Dense, Embedding, GlobalAveragePooling1D
+from tensorflow.keras.preprocessing.text import Tokenizer
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+```
+
+### 2. Load Dataset:
+
+```python3
+data = pd.read_csv('IMDB Dataset.csv')
+print(data.head())
+```
+
+### 3. Exploratory Data Analysis (EDA):
+
+```python3
+print("Shape:", data.shape)
+print("\nMissing Values:\n", data.isnull().sum())
+print("\nClass Distribution:\n", data['sentiment'].value_counts())
+
+# Visualize class distribution
+sns.countplot(x='sentiment', data=data)
+plt.title('Sentiment Class Distribution')
+plt.show()
+
+# Sample reviews
+print("\nSample positive review:\n", data[data['sentiment'] == 'positive']['review'].iloc[0][:300])
+print("\nSample negative review:\n", data[data['sentiment'] == 'negative']['review'].iloc[0][:300])
+```
+
+### 4. Data Cleaning - Strip HTML Tags:
+
+```python3
+def clean_text(text):
+    text = re.sub(r'<.*?>', '', text)   # remove HTML tags like <br />
+    text = text.lower().strip()         # lowercase and trim whitespace
+    return text
+
+data['review'] = data['review'].apply(clean_text)
+print("Sample cleaned review:\n", data['review'].iloc[0][:300])
+```
+
+### 5. Encode Labels and Separate Features:
+
+```python3
+label_encoder = LabelEncoder()
+data['sentiment'] = label_encoder.fit_transform(data['sentiment'])  # positive=1, negative=0
+
+X = data['review'].values    # input: review text
+y = data['sentiment'].values # output: 0 or 1
+```
+
+### 6. Tokenize and Pad Text Sequences:
+
+```python3
+vocab_size = 10000   # keep only top 10,000 most frequent words
+max_length = 200     # truncate/pad all reviews to 200 words
+
+tokenizer = Tokenizer(num_words=vocab_size, oov_token='<OOV>')  # <OOV> handles unknown words
+tokenizer.fit_on_texts(X)                        # build word index from training text
+
+sequences = tokenizer.texts_to_sequences(X)      # convert each word to its integer index
+padded_sequences = pad_sequences(sequences, maxlen=max_length,
+                                 padding='post', truncating='post')  # pad/truncate to fixed length
+```
+
+### 7. Split into Training and Testing Sets:
+
+```python3
+X_train, X_test, y_train, y_test = train_test_split(padded_sequences, y, test_size=0.2, random_state=42)
+```
+
+### 8. Build the Neural Network Model:
+
+```python3
+model = Sequential()
+model.add(Embedding(vocab_size, 16))        # maps each word index to a 16-dim vector
+model.add(GlobalAveragePooling1D())          # averages all word vectors into one vector
+model.add(Dense(24, activation='relu'))      # hidden layer: 24 neurons
+model.add(Dense(1, activation='sigmoid'))    # output: probability between 0 and 1 (binary)
+
+model.summary()
+```
+
+### 9. Compile the Model:
+
+```python3
+# binary_crossentropy: standard loss for binary classification; sigmoid output
+model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
+```
+
+### 10. Train the Model:
+
+```python3
+history = model.fit(X_train, y_train, epochs=10, batch_size=32, validation_split=0.2)
+```
+
+### 11. Evaluate the Model on Test Data:
+
+```python3
+loss, accuracy = model.evaluate(X_test, y_test)
+print(f"Test Loss: {loss:.4f}")
+print(f"Test Accuracy: {accuracy*100:.2f}%")
+```
+
+### 12. Plot Training vs Validation Accuracy:
+
+```python3
+plt.plot(history.history['accuracy'], label='Training Accuracy')
+plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
+plt.title('Model Accuracy Over Epochs')
+plt.ylabel('Accuracy')
+plt.xlabel('Epoch')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 13. Plot Training vs Validation Loss:
+
+```python3
+plt.plot(history.history['loss'], label='Training Loss')
+plt.plot(history.history['val_loss'], label='Validation Loss')
+plt.title('Model Loss Over Epochs')
+plt.ylabel('Loss')
+plt.xlabel('Epoch')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 14. Confusion Matrix and Classification Report:
+
+```python3
+y_pred = (model.predict(X_test) > 0.5).astype(int)  # threshold 0.5: prob > 0.5 = positive
+
+cm = confusion_matrix(y_test, y_pred)
+sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
+            xticklabels=['Negative', 'Positive'],
+            yticklabels=['Negative', 'Positive'])
+plt.title('Confusion Matrix')
+plt.ylabel('Actual')
+plt.xlabel('Predicted')
+plt.show()
+
+print("\nClassification Report:\n", classification_report(y_test, y_pred, target_names=['Negative', 'Positive']))
+```
+
+---
+
+## Miscellaneous
+
+- [Dataset source](https://www.kaggle.com/datasets/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews)
+
+---
+

Codes/Code-3a.md

@@ -0,0 +1,240 @@
+# Practical-3a (Convolutional Neural Network - Plant Diseases)
+
+Problem Statement: Convolutional Neural Network (CNN): Use any dataset of plant disease and design a plant disease detection system using CNN.
+
+> [!NOTE]
+> Download dataset directly from [source](https://www.kaggle.com/datasets/vipoooool/new-plant-diseases-dataset/data).
+> Haven't added it to the `/Datasets` directory due to its large size.
+> tbh the dataset doesn't really matter in this case, you just need to ensure dataset directory contains `train` and `valid` sub-directories.
+> Refer the above dataset to understand the required directory structure.
+
+---
+
+## Pre-requisities
+
+1. Install packages using `pip`: `pip install tensorflow keras numpy opencv-python matplotlib seaborn scikit-learn` (`tensorflow` requires Python 3.9 - 3.12)
+2. Download and unzip the dataset in the same directory as the Jupyter notebook.
+3. Ensure your unzipped dataset has the required directory structure:
+
+```shell
+New Plant Diseases Dataset(Augmented)/
+├── train
+│   ├── Apple___Apple_scab
+│   ├── Apple___Black_rot
+│   ├── Apple___Cedar_apple_rust
+├── valid
+│   ├── Apple___Apple_scab
+│   ├── Apple___Black_rot
+│   ├── Apple___Cedar_apple_rust
+```
+
+## Steps
+
+1. Import Libraries
+2. Load Dataset
+3. Exploratory Data Analysis (EDA)
+4. Split into Training and Testing Sets
+5. Build the CNN Model
+6. Compile the Model
+7. Train the Model
+8. Evaluate the Model on Test Data
+9. Plot Training vs Validation Accuracy
+10. Plot Training vs Validation Loss
+11. Confusion Matrix and Classification Report
+
+---
+
+## Code
+
+### 1. Import Libraries:
+
+```python3
+import os
+import numpy as np
+import cv2
+import matplotlib.pyplot as plt
+import seaborn as sns
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import confusion_matrix, classification_report
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Flatten, Dense, Dropout
+from tensorflow.keras.utils import to_categorical
+```
+
+### 2. Load Dataset:
+
+```python3
+data = []
+labels = []
+
+# Path to dataset folder containing one subfolder per disease class
+path = './New Plant Diseases Dataset(Augmented)/train/'
+categories = sorted(os.listdir(path))  # sort for consistent label ordering
+
+# Map each category name to a numeric index
+label_dict = {category: idx for idx, category in enumerate(categories)}
+print("Classes found:", len(categories))
+
+max_per_class = 200  # cap images per class to avoid RAM overflow on large datasets
+
+for category in categories:
+    folder = os.path.join(path, category)
+    count = 0
+    for img_name in os.listdir(folder):
+        if count >= max_per_class:
+            break
+        img_path = os.path.join(folder, img_name)
+        img_array = cv2.imread(img_path)
+        if img_array is not None:                          # skip unreadable files
+            img_array = cv2.resize(img_array, (64, 64))   # resize to fixed 64x64 pixels
+            data.append(img_array)
+            labels.append(label_dict[category])
+            count += 1
+
+data = np.array(data) / 255.0   # normalize pixel values from [0,255] to [0,1]
+labels = np.array(labels)
+print("Dataset shape:", data.shape)
+print("Labels shape:", labels.shape)
+```
+
+### 3. Exploratory Data Analysis (EDA):
+
+```python3
+print("Total images:", len(data))
+print("Image shape:", data[0].shape)
+print("Number of classes:", len(categories))
+
+# Class distribution bar chart
+class_counts = {cat: int((labels == idx).sum()) for cat, idx in label_dict.items()}
+plt.figure(figsize=(14, 5))
+plt.bar(class_counts.keys(), class_counts.values())
+plt.xticks(rotation=90)
+plt.title("Number of Images per Disease Class")
+plt.xlabel("Class")
+plt.ylabel("Count")
+plt.tight_layout()
+plt.show()
+
+# Sample images from first 5 classes
+plt.figure(figsize=(15, 3))
+for i, category in enumerate(categories[:5]):
+    idx = np.where(labels == label_dict[category])[0][0]  # index of first image in class
+    plt.subplot(1, 5, i + 1)
+    plt.imshow(cv2.cvtColor((data[idx] * 255).astype(np.uint8), cv2.COLOR_BGR2RGB))
+    plt.title(category[:15], fontsize=8)
+    plt.axis('off')
+plt.suptitle("Sample Images per Class")
+plt.show()
+```
+
+### 4. Split into Training and Testing Sets:
+
+```python3
+X_train, X_test, y_train, y_test = train_test_split(data, labels, test_size=0.2, random_state=42)
+
+num_classes = len(categories)
+# One-hot encode labels: e.g. class 2 of 5 → [0, 0, 1, 0, 0]
+y_train = to_categorical(y_train, num_classes)
+y_test  = to_categorical(y_test,  num_classes)
+print("Train samples:", X_train.shape[0])
+print("Test samples: ", X_test.shape[0])
+```
+
+### 5. Build the CNN Model:
+
+```python3
+model = Sequential()
+
+model.add(Input(shape=(64, 64, 3)))               # input: 64x64 RGB image
+model.add(Conv2D(32, (3, 3), activation='relu'))  # 32 filters, detect basic features
+model.add(MaxPooling2D(2, 2))                      # downsample by 2x
+
+model.add(Conv2D(64, (3, 3), activation='relu'))  # 64 filters, detect complex features
+model.add(MaxPooling2D(2, 2))
+
+model.add(Flatten())                              # convert 2D feature maps to 1D vector
+
+model.add(Dense(128, activation='relu'))          # fully connected layer
+model.add(Dropout(0.5))                           # randomly drop 50% neurons to reduce overfitting
+
+model.add(Dense(num_classes, activation='softmax'))  # output: probability for each class
+
+model.summary()
+```
+
+### 6. Compile the Model:
+
+```python3
+# categorical_crossentropy: standard loss for multi-class classification with one-hot labels
+model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
+```
+
+### 7. Train the Model:
+
+```python3
+history = model.fit(X_train, y_train, epochs=10, batch_size=32, validation_split=0.2)
+```
+
+### 8. Evaluate the Model on Test Data:
+
+```python3
+loss, accuracy = model.evaluate(X_test, y_test)
+print(f"Test Loss: {loss:.4f}")
+print(f"Test Accuracy: {accuracy*100:.2f}%")
+```
+
+### 9. Plot Training vs Validation Accuracy:
+
+```python3
+plt.plot(history.history['accuracy'], label='Training Accuracy')
+plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
+plt.title('CNN Model Accuracy Over Epochs')
+plt.xlabel('Epoch')
+plt.ylabel('Accuracy')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 10. Plot Training vs Validation Loss:
+
+```python3
+plt.plot(history.history['loss'], label='Training Loss')
+plt.plot(history.history['val_loss'], label='Validation Loss')
+plt.title('CNN Model Loss Over Epochs')
+plt.xlabel('Epoch')
+plt.ylabel('Loss')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 11. Confusion Matrix and Classification Report:
+
+```python3
+y_pred = np.argmax(model.predict(X_test), axis=1)  # predicted class index
+y_true = np.argmax(y_test, axis=1)                  # actual class index (from one-hot)
+
+cm = confusion_matrix(y_true, y_pred)
+plt.figure(figsize=(14, 12))
+sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
+            xticklabels=categories, yticklabels=categories)
+plt.title('Confusion Matrix')
+plt.ylabel('Actual')
+plt.xlabel('Predicted')
+plt.xticks(rotation=90)
+plt.tight_layout()
+plt.show()
+
+print("\nClassification Report:\n")
+print(classification_report(y_true, y_pred, target_names=categories))
+```
+
+---
+
+## Miscellaneous
+
+- [Dataset source](https://www.kaggle.com/datasets/vipoooool/new-plant-diseases-dataset)
+
+---
+

Codes/Code-3b.md

@@ -0,0 +1,233 @@
+# Practical-3b (Convolutional Neural Network - MNIST Fashion Dataset)
+
+Problem Statement: Convolutional Neural Network (CNN): Use MNIST Fashion Dataset and create a classifier to classify fashion clothing into categories.
+
+> [!NOTE]
+> Download dataset directly from [source](https://www.kaggle.com/datasets/zalando-research/fashionmnist).
+> Dataset available in [Datasets](../Datasets/fashionmnist.zip) directory.
+> In the code, dataset is downloaded directly from Keras/TensorFlow in 2nd step (Load Dataset)
+
+---
+
+## Pre-requisities
+
+1. Install packages using `pip`: `pip install tensorflow keras numpy matplotlib seaborn scikit-learn` (`tensorflow` requires Python 3.9 - 3.12)
+
+## Steps
+
+---
+
+## Code
+
+### 1. Import Libraries:
+
+```python3
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+import tensorflow as tf
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Input, Conv2D, AvgPool2D, GlobalAveragePooling2D, Dense
+from tensorflow.keras.utils import to_categorical
+from sklearn.metrics import confusion_matrix, classification_report
+```
+
+### 2. Load Dataset:
+
+```python3
+# Fashion MNIST is built into Keras, downloads automatically on first run
+(X_train, y_train), (X_test, y_test) = tf.keras.datasets.fashion_mnist.load_data()
+
+'''
+import numpy as np
+import gzip
+import os
+
+def load_fashion_mnist(path):
+    """Load Fashion MNIST from local .gz files (Kaggle Zalando format)."""
+    files = {
+        'X_train': 'train-images-idx3-ubyte.gz',
+        'y_train': 'train-labels-idx1-ubyte.gz',
+        'X_test':  't10k-images-idx3-ubyte.gz',
+        'y_test':  't10k-labels-idx1-ubyte.gz',
+    }
+
+    with gzip.open(os.path.join(path, files['X_train'])) as f:
+        X_train = np.frombuffer(f.read(), np.uint8, offset=16).reshape(-1, 28, 28)
+    with gzip.open(os.path.join(path, files['y_train'])) as f:
+        y_train = np.frombuffer(f.read(), np.uint8, offset=8)
+    with gzip.open(os.path.join(path, files['X_test'])) as f:
+        X_test = np.frombuffer(f.read(), np.uint8, offset=16).reshape(-1, 28, 28)
+    with gzip.open(os.path.join(path, files['y_test'])) as f:
+        y_test = np.frombuffer(f.read(), np.uint8, offset=8)
+
+    return (X_train, y_train), (X_test, y_test)
+
+# Replace the Keras load line with:
+(X_train, y_train), (X_test, y_test) = load_fashion_mnist('./fashion-mnist/')
+'''
+
+print("Training set shape:", X_train.shape)   # (60000, 28, 28)
+print("Test set shape:    ", X_test.shape)    # (10000, 28, 28)
+print("Classes:", np.unique(y_train))
+```
+
+### 3. Exploratory Data Analysis (EDA):
+
+```python3
+class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
+               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']
+
+# Class distribution
+unique, counts = np.unique(y_train, return_counts=True)
+plt.figure(figsize=(10, 4))
+plt.bar([class_names[i] for i in unique], counts)
+plt.xticks(rotation=45, ha='right')
+plt.title("Training Set Class Distribution")
+plt.ylabel("Count")
+plt.tight_layout()
+plt.show()
+
+# Sample images (one per class)
+plt.figure(figsize=(15, 3))
+for i, cls in enumerate(class_names):
+    idx = np.where(y_train == i)[0][0]   # index of first image for this class
+    plt.subplot(1, 10, i + 1)
+    plt.imshow(X_train[idx], cmap='gray')
+    plt.title(cls, fontsize=7)
+    plt.axis('off')
+plt.suptitle("Sample Image per Class")
+plt.tight_layout()
+plt.show()
+```
+
+### 4. Preprocess Data:
+
+```python3
+# Reshape to add channel dimension: (samples, 28, 28) -> (samples, 28, 28, 1)
+X_train = X_train.reshape(-1, 28, 28, 1).astype('float32') / 255.0  # normalize to [0,1]
+X_test  = X_test.reshape(-1, 28, 28, 1).astype('float32') / 255.0
+
+# One-hot encode labels: e.g. class 3 of 10 -> [0,0,0,1,0,0,0,0,0,0]
+y_train_cat = to_categorical(y_train, num_classes=10)
+y_test_cat  = to_categorical(y_test,  num_classes=10)
+
+print("X_train shape:", X_train.shape)
+print("y_train_cat shape:", y_train_cat.shape)
+```
+
+### 5. Build the CNN Model:
+
+```python3
+model = Sequential()
+
+model.add(Input(shape=(28, 28, 1)))                                          # input: 28x28 grayscale image
+model.add(Conv2D(64, kernel_size=(3, 3), activation='relu', padding='same')) # 64 filters, extract features
+model.add(AvgPool2D(pool_size=(2, 2)))                                        # downsample to 14x14
+
+model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', padding='same')) # 32 filters, refine features
+model.add(AvgPool2D(pool_size=(2, 2)))                                        # downsample to 7x7
+
+model.add(GlobalAveragePooling2D())                                           # average each feature map to single value
+model.add(Dense(10, activation='softmax'))                                    # output: probability for each of 10 classes
+
+model.summary()
+```
+
+### 6. Compile the Model:
+
+```python3
+# categorical_crossentropy: standard loss for multi-class one-hot classification
+model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
+```
+
+### 7. Train the Model:
+
+```python3
+# validation_data uses test set to monitor performance after each epoch
+history = model.fit(X_train, y_train_cat, epochs=10, validation_data=(X_test, y_test_cat))
+```
+
+### 8. Evaluate the Model on Test Data:
+
+```python3
+loss, accuracy = model.evaluate(X_test, y_test_cat)
+print(f"Test Loss:     {loss:.4f}")
+print(f"Test Accuracy: {accuracy*100:.2f}%")
+```
+
+### 9. Plot Training vs Validation Accuracy:
+
+```python3
+plt.plot(history.history['accuracy'], label='Training Accuracy')
+plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
+plt.title('Model Accuracy Over Epochs')
+plt.xlabel('Epoch')
+plt.ylabel('Accuracy')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 10. Plot Training vs Validation Loss:
+
+```python3
+plt.plot(history.history['loss'], label='Training Loss')
+plt.plot(history.history['val_loss'], label='Validation Loss')
+plt.title('Model Loss Over Epochs')
+plt.xlabel('Epoch')
+plt.ylabel('Loss')
+plt.legend()
+plt.grid(True)
+plt.show()
+```
+
+### 11. Confusion Matrix and Classification Report:
+
+```python3
+y_pred = np.argmax(model.predict(X_test), axis=1)  # predicted class index
+
+cm = confusion_matrix(y_test, y_pred)
+plt.figure(figsize=(10, 8))
+sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
+            xticklabels=class_names, yticklabels=class_names)
+plt.title('Confusion Matrix')
+plt.ylabel('Actual')
+plt.xlabel('Predicted')
+plt.xticks(rotation=45, ha='right')
+plt.tight_layout()
+plt.show()
+
+print("\nClassification Report:\n")
+print(classification_report(y_test, y_pred, target_names=class_names))
+```
+
+### 12. Visualize Sample Predictions:
+
+```python3
+# batch predict all test images, then pick 10 random ones to display
+all_preds = np.argmax(model.predict(X_test), axis=1)
+random_indices = np.random.choice(len(X_test), 10, replace=False)
+
+plt.figure(figsize=(20, 4))
+for i, idx in enumerate(random_indices):
+    plt.subplot(2, 5, i + 1)
+    plt.imshow(X_test[idx].reshape(28, 28), cmap='gray')
+    predicted = class_names[all_preds[idx]]
+    actual    = class_names[y_test[idx]]
+    color = 'green' if predicted == actual else 'red'   # green = correct, red = wrong
+    plt.title(f"P: {predicted}\nA: {actual}", fontsize=8, color=color)
+    plt.axis('off')
+plt.suptitle("Sample Predictions (Green=Correct, Red=Wrong)")
+plt.tight_layout()
+plt.show()
+```
+
+---
+
+## Miscellaneous
+
+- [Dataset source](https://www.kaggle.com/datasets/zalando-research/fashionmnist)
+
+---
+

Codes/Code-4.md

@@ -0,0 +1,276 @@
+# Practical-4 (Recurrent Neural Network - Google Stock Price Dataset)
+
+Problem Statement: Recurrent neural network (RNN): Use the Google stock prices dataset and design a time series analysis and prediction system using RNN.
+
+> [!NOTE]
+> Dataset available in [Datasets](../Datasets/GOOG.csv) directory.
+> In the code, dataset is downloaded directly from Keras/TensorFlow in 2nd step (Load Dataset)
+
+---
+
+## Pre-requisities
+
+1. Install packages using `pip`: `pip install tensorflow keras numpy pandas matplotlib scikit-learn yfinance` (`tensorflow` requires Python 3.9 - 3.12)
+
+## Steps
+
+1. Import Libraries
+2. Load Dataset
+3. Exploratory Data Analysis (EDA)
+4. Visualize Closing Price Over Time
+5. Preprocess Data - Normalize Closing Price
+6. Create Sequences for RNN Input
+7. Build the RNN Model
+8. Train the Model
+9. Plot Training vs Validation Loss
+10. Make Predictions and Inverse Scale
+11. Evaluate the Model
+12. Plot Actual vs Predicted Stock Price
+13. Forecast Next 30 Days
+
+---
+
+## Code
+
+### 1. Import Libraries:
+
+```python3
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+import yfinance as yf
+from sklearn.preprocessing import MinMaxScaler
+from sklearn.metrics import mean_squared_error, mean_absolute_error
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Input, Dense, SimpleRNN, Dropout
+from tensorflow.keras.callbacks import EarlyStopping
+```
+
+### 2. Load Dataset:
+
+```python3
+# Downloads GOOGL stock data from Yahoo Finance for the given date range
+ticker = "GOOGL"
+df = yf.download(ticker, start="2018-01-01", end="2024-01-01")
+
+# --- Offline alternative (comment out the yf.download above and use this instead if using local dataset) ---
+# df = pd.read_csv('GOOGL.csv', index_col='Date', parse_dates=True)
+# df = df.sort_index()  # ensure chronological order
+
+# yfinance returns MultiIndex columns — flatten to single level
+df.columns = df.columns.get_level_values(0)
+
+print(f"Dataset Shape: {df.shape}")
+print(f"Date Range: {df.index.min().date()} to {df.index.max().date()}")
+print(df.head())
+```
+
+### 3. Exploratory Data Analysis (EDA):
+
+```python3
+print("=== Dataset Info ===")
+print(df.info())
+print("\n=== Statistical Summary ===")
+print(df.describe())
+print("\n=== Missing Values ===")
+print(df.isnull().sum())
+```
+
+### 4. Visualize Closing Price Over Time:
+
+```python3
+plt.figure(figsize=(16, 6))
+plt.plot(df.index, df['Close'], color='steelblue', linewidth=1.5, label='Close Price')
+plt.title('Google (GOOGL) Stock Closing Price (2018–2024)')
+plt.xlabel('Date')
+plt.ylabel('Price (USD)')
+plt.legend()
+plt.grid(alpha=0.3)
+plt.tight_layout()
+plt.show()
+```
+
+### 5. Preprocess Data - Normalize Closing Price:
+
+```python3
+data = df[['Close']].values   # use only Close price for prediction
+
+scaler = MinMaxScaler(feature_range=(0, 1))
+data_scaled = scaler.fit_transform(data)  # scale values to [0, 1]
+
+print(f"Original data range: [{data.min():.2f}, {data.max():.2f}]")
+print(f"Scaled data range:   [{data_scaled.min():.4f}, {data_scaled.max():.4f}]")
+print(f"Total data points:   {len(data_scaled)}")
+```
+
+### 6. Create Sequences for RNN Input:
+
+```python3
+def create_sequences(data, time_steps=60):
+    X, y = [], []
+    for i in range(time_steps, len(data)):
+        X.append(data[i - time_steps:i, 0])  # window of past `time_steps` days
+        y.append(data[i, 0])                  # next day's price
+    return np.array(X), np.array(y)
+
+TIME_STEPS = 60  # use past 60 days to predict the next day
+
+# 80/20 train-test split (manual, to preserve time order)
+train_size = int(len(data_scaled) * 0.80)
+train_data = data_scaled[:train_size]
+test_data  = data_scaled[train_size - TIME_STEPS:]  # overlap ensures test sequences start correctly
+
+X_train, y_train = create_sequences(train_data, TIME_STEPS)
+X_test,  y_test  = create_sequences(test_data,  TIME_STEPS)
+
+# Reshape to [samples, time_steps, features] — required format for RNN layers
+X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], 1))
+X_test  = X_test.reshape((X_test.shape[0],   X_test.shape[1],  1))
+
+print(f"Training samples: {X_train.shape}")
+print(f"Testing samples:  {X_test.shape}")
+```
+
+### 7. Build the RNN Model:
+
+```python3
+model = Sequential()
+
+model.add(Input(shape=(TIME_STEPS, 1)))                               # input: sequence of 60 days
+model.add(SimpleRNN(units=64, return_sequences=True))                 # first RNN layer, passes output to next
+model.add(Dropout(0.2))                                               # drop 20% neurons to reduce overfitting
+model.add(SimpleRNN(units=64, return_sequences=False))                # second RNN layer, outputs single vector
+model.add(Dropout(0.2))
+model.add(Dense(units=32, activation='relu'))                         # fully connected layer
+model.add(Dense(units=1))                                             # output: single predicted price
+
+model.compile(optimizer='adam', loss='mean_squared_error', metrics=['mae'])
+model.summary()
+```
+
+### 8. Train the Model:
+
+```python3
+# EarlyStopping stops training if val_loss doesn't improve for 10 consecutive epochs
+early_stop = EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)
+
+history = model.fit(
+    X_train, y_train,
+    epochs=60,
+    batch_size=32,
+    validation_split=0.1,   # use 10% of training data for validation
+    callbacks=[early_stop],
+    verbose=1
+)
+print(f"\nTraining stopped at epoch: {len(history.history['loss'])}")
+```
+
+### 9. Plot Training vs Validation Loss:
+
+```python3
+plt.plot(history.history['loss'], label='Train Loss', color='royalblue')
+plt.plot(history.history['val_loss'], label='Val Loss', color='tomato')
+plt.title('Model Training Loss Over Epochs')
+plt.xlabel('Epoch')
+plt.ylabel('MSE Loss')
+plt.legend()
+plt.grid(alpha=0.3)
+plt.tight_layout()
+plt.show()
+```
+
+### 10. Make Predictions and Inverse Scale:
+
+```python3
+y_pred_scaled = model.predict(X_test)
+
+# Convert scaled predictions back to original USD price range
+y_pred   = scaler.inverse_transform(y_pred_scaled)
+y_actual = scaler.inverse_transform(y_test.reshape(-1, 1))
+
+print(f"Sample predictions (first 5): {y_pred[:5].flatten().round(2)}")
+print(f"Actual values      (first 5): {y_actual[:5].flatten().round(2)}")
+```
+
+### 11. Evaluate the Model:
+
+```python3
+mse  = mean_squared_error(y_actual, y_pred)
+rmse = np.sqrt(mse)
+mae  = mean_absolute_error(y_actual, y_pred)
+mape = np.mean(np.abs((y_actual - y_pred) / y_actual)) * 100  # mean absolute percentage error
+
+print("=" * 40)
+print("     MODEL EVALUATION METRICS")
+print("=" * 40)
+print(f"  MSE  : {mse:.4f}")
+print(f"  RMSE : {rmse:.4f}")
+print(f"  MAE  : {mae:.4f}")
+print(f"  MAPE : {mape:.2f}%")
+print("=" * 40)
+```
+
+### 12. Plot Actual vs Predicted Stock Price:
+
+```python3
+test_dates = df.index[train_size:]   # align dates with test predictions
+
+plt.figure(figsize=(16, 6))
+plt.plot(test_dates, y_actual, label='Actual Price',    color='steelblue', linewidth=1.5)
+plt.plot(test_dates, y_pred,   label='Predicted Price', color='tomato',    linewidth=1.5, linestyle='--')
+plt.title('Google Stock Price: Actual vs Predicted (RNN)')
+plt.xlabel('Date')
+plt.ylabel('Price (USD)')
+plt.legend()
+plt.grid(alpha=0.3)
+plt.tight_layout()
+plt.show()
+```
+
+### 13. Forecast Next 30 Days:
+
+```python3
+n_future = 30  # number of future days to predict
+
+# Seed the forecast with the last TIME_STEPS days of known data
+future_input       = data_scaled[-TIME_STEPS:].reshape(1, TIME_STEPS, 1)
+future_predictions = []
+
+for _ in range(n_future):
+    pred = model.predict(future_input, verbose=0)
+    future_predictions.append(pred[0, 0])
+    # Slide the window: drop oldest day, append new prediction
+    future_input = np.append(future_input[:, 1:, :], pred.reshape(1, 1, 1), axis=1)
+
+# Inverse scale forecasted prices back to USD
+future_prices = scaler.inverse_transform(np.array(future_predictions).reshape(-1, 1))
+
+# Generate business day dates starting from the day after last known date
+last_date    = df.index[-1]
+future_dates = pd.date_range(start=last_date + pd.Timedelta(days=1), periods=n_future, freq='B')
+
+plt.figure(figsize=(16, 6))
+plt.plot(df.index[-120:], scaler.inverse_transform(data_scaled[-120:]),
+         label='Historical', color='steelblue', linewidth=1.5)
+plt.plot(future_dates, future_prices,
+         label='30-Day Forecast', color='orange', linewidth=1.5)
+plt.axvline(x=last_date, color='gray', linestyle='--', label='Forecast Start')
+plt.title('Google Stock — 30-Day Future Price Forecast (RNN)')
+plt.xlabel('Date')
+plt.ylabel('Price (USD)')
+plt.legend()
+plt.grid(alpha=0.3)
+plt.tight_layout()
+plt.show()
+
+print(f"\nForecasted price range: {future_prices.min():.2f} USD - {future_prices.max():.2f} USD")
+```
+
+---
+
+## Miscellaneous
+
+- [Dataset source](https://www.kaggle.com/datasets/henryshan/google-stock-price)
+
+---
+

README.md

@@ -8,6 +8,30 @@ This repository gathers comprehensive material for the SPPU Computer Engineering
 
 ### Notes
 
+### Codes
+
+1. [Code-1 (Linear Regression using Deep Neural Network)](Codes/Code-1.md)
+2. [Code-2b (Classification using Deep Neural Network)](Codes/Code-2b.md)
+3. [Code-3a (Convolutional Neural Network - Plant Diseases)](Codes/Code-3a.md)
+4. [Code-3b (Convolutional Neural Network - MNIST Fashion Dataset)](Codes/Code-3b.md)
+5. [Code-4 (Recurrent Neural Network - Google Stock Price Dataset)](Codes/Code-4.md)
+
+### Jupyter Notebooks
+
+1. [Notebook-1 (Linear Regression using Deep Neural Network)](Notebooks/Notebook-1.ipynb)
+2. [Notebook-2b (Classification using Deep Neural Network)](Notebooks/Notebook-2b.ipynb)
+3. [Notebook-3a (Convolutional Neural Network - Plant Diseases)](Notebooks/Notebook-3a.ipynb)
+4. [Notebook-3b (Convolutional Neural Network - MNIST Fashion Dataset)](Notebooks/Notebook-3b.ipynb)
+5. [Notebook-4 (Recurrent Neural Network - Google Stock Price Dataset)](Notebooks/Notebook-4.ipynb)
+
+
+### Datasets
+
+1. [Dataset for Practical-1 (Boston House Price)](Datasets/boston.csv)
+2. [Dataset for Practical-2b (IMDB Reviews)](Datasets/IMDB%20Dataset.csv)
+3. [Dataset for Practical-3b (MNIST Fashion)](Datasets/fashionmnist.zip)
+4. [Dataset for Practical-4 (Google Stock Price)](Datasets/GOOG.csv)
+
 ### Assignments
 
 - [Questions - Assignment 1 and 2](Assignments/DL%20-%20Assignments-1+2%20%28Questions%29.pdf)