I am working on a personal Machine-Learning (ML) project to predict weather on Jupyter Notebook. Once I am done with it, I will start working on converting it into a Flask app. I have completed my code on Jupyter Notebook. All the codes are working fine. However, I am not sure if everything is done correctly. Would you please review my code on GitHub and inform me if I am doing it the right way? https://github.com/SteveAustin583/w...ml-stackexchange-feedback-implemented-3.ipynb Here is my code: # %% [markdown] # # Seattle Weather Category Prediction - Jupyter Notebook # # ## This notebook aims to predict the weather category (e.g., sun, rain, snow) # # based on other meteorological features. It follows a structure similar to # # the provided Kaggle example, using Gaussian Naive Bayes. # # ## Steps: # (Original numbered steps removed as per feedback, structure implied by headers) # - Load and Explore Data # - Visualize Data # - Preprocess Data (Label Encoding for target, add lagged features) # - Train and Evaluate Models # - Conduct Ablation Study # - Save the Model and Label Encoder for Flask App # %% [markdown] # ## Setup and Load Data # %% import numpy as np import pandas as pd import seaborn as sns import matplotlib.pyplot as plt import io # To load CSV from string in some environments (not strictly used by pd.read_csv here) from sklearn.model_selection import train_test_split from sklearn.naive_bayes import GaussianNB from sklearn.linear_model import LogisticRegression from sklearn.svm import SVC from sklearn.metrics import accuracy_score, classification_report, confusion_matrix from sklearn.preprocessing import LabelEncoder import joblib # %% # Display plots inline %matplotlib inline # Set some display options for Pandas pd.set_option('display.max_columns', None) pd.set_option('display.width', 1000) # %% [markdown] # ### Helper Functions # %% def print_header(message): """Prints a formatted header message.""" print(f"\n{message}:") def load_and_preprocess_data(file_path): """Loads the dataset and performs initial date conversion.""" df = pd.read_csv(file_path) print_header("Dataset loaded successfully") df['date'] = pd.to_datetime(df['date']) # Convert date column to datetime return df def create_visualization_df(dataframe): """Creates a copy of the dataframe for visualization and extracts year/month.""" df_vis = dataframe.copy() df_vis['year'] = df_vis['date'].dt.year df_vis['month'] = df_vis['date'].dt.month return df_vis # --- Feedback Implementation: Helper Defaults and Splitting --- _DEFAULT_FEATURES_TO_LAG = ("precipitation", "temp_max", "temp_min", "wind", "weather_encoded") _DEFAULT_TEMP_FEATURES_FOR_DELTA = ("temp_max", "temp_min") def _create_lag_features(df_input, features_to_lag=_DEFAULT_FEATURES_TO_LAG, lag_period=1): """ Adds lagged versions of specified features. Sorts by 'date' column if present before lagging. """ df_out = df_input.copy() if 'date' in df_out.columns: df_out = df_out.sort_values(by='date').reset_index(drop=True) # Sort and reset index else: print("Warning: 'date' column not present for sorting in _create_lag_features. Assuming pre-sorted data for lag features.") for feature in features_to_lag: if feature in df_out.columns: df_out[f'{feature}_lag{lag_period}'] = df_out[feature].shift(lag_period) else: print(f"Warning: Feature '{feature}' not found in DataFrame for lagging.") return df_out # Returns DataFrame with NaNs from shifting def _create_delta_features(df_input, temp_features_for_delta=_DEFAULT_TEMP_FEATURES_FOR_DELTA, lag_period=1): """ Adds delta (difference) features for specified temperature-like features. Assumes df_input already contains the necessary current and lagged features. """ df_out = df_input.copy() for temp_feature in temp_features_for_delta: current_feature_name = temp_feature lagged_feature_name = f'{temp_feature}_lag{lag_period}' if current_feature_name in df_out.columns and lagged_feature_name in df_out.columns: df_out[f'delta_{temp_feature}'] = df_out[current_feature_name] - df_out[lagged_feature_name] else: if current_feature_name not in df_out.columns: print(f"Warning: Current feature {current_feature_name} not found for delta calculation.") if lagged_feature_name not in df_out.columns: print(f"Warning: Lagged feature {lagged_feature_name} not found for delta calculation of {current_feature_name}.") return df_out def train_and_evaluate_model(model, X_train, y_train, X_test, y_test, target_names, model_name="Model"): """Trains a model and prints evaluation metrics.""" print_header(f"--- {model_name} Training and Evaluation ---") model.fit(X_train, y_train) y_pred = model.predict(X_test) accuracy = accuracy_score(y_test, y_pred) conf_matrix = confusion_matrix(y_test, y_pred) classification_rep = classification_report(y_test, y_pred, target_names=target_names, zero_division=0) print(f"Accuracy: {accuracy:.4f}") print_header("Confusion Matrix") plt.figure(figsize=(8, 6)) sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='Blues', xticklabels=target_names, yticklabels=target_names) plt.xlabel('Predicted Label') plt.ylabel('True Label') plt.title(f'Confusion Matrix - {model_name}') plt.show() print_header("Classification Report") print(classification_rep) return model, accuracy # %% # --- Load the dataset --- file_path = 'seattle-weather.csv' df = load_and_preprocess_data(file_path) df.head() # %% [markdown] # ## Initial Data Exploration & Visualization # %% print_header("Dataset Info") df.info() # %% print_header("Statistical Summary") print(df.describe()) # %% print_header("Missing Values Check") print(df.isnull().sum()) print(f"Any NA values present: {df.isna().sum().any()}") # %% print_header("Duplicate Rows Check") print(f"Number of duplicated rows: {df.duplicated().sum()}") # %% print_header("Day with Minimum temp_min") print(df[df['temp_min']==min(df.temp_min)]) # %% print_header("Day with Maximum temp_max") print(df[df['temp_max']==max(df.temp_max)]) # %% t_min_overall = pd.concat([df['temp_min'], df['temp_max']]).min() t_max_overall = pd.concat([df['temp_min'], df['temp_max']]).max() bins = np.arange(np.floor(t_min_overall), np.ceil(t_max_overall) + 1, 1) plt.figure(figsize=(12,6)) sns.histplot(data=df, x='temp_max', bins=bins, kde=True, palette="viridis") # Using palette instead of color for histplot plt.title('Distribution of Maximum Temperature') plt.xlabel('Max Temperature (°C)') plt.ylabel('Frequency') plt.xlim(bins.min(), bins.max()) plt.xticks(bins[::2]) plt.show() # %% plt.figure(figsize=(12,6)) sns.histplot(data=df, x='temp_min', bins=bins, kde=True, palette="viridis") # Using palette instead of color for histplot plt.title('Distribution of Minimum Temperature') plt.xlabel('Min Temperature (°C)') plt.ylabel('Frequency') plt.xlim(bins.min(), bins.max()) plt.xticks(bins[::2]) plt.show() # %% [markdown] # ### FacetGrid Visualizations (Month vs. Weather Variables by Year) # %% df_vis = create_visualization_df(df) # %% g = sns.FacetGrid(df_vis, col='year', col_wrap=4, height=3.5, aspect=1.2) g.map(sns.lineplot, 'month', 'temp_max', errorbar=None) g.set_axis_labels('Month', 'Max Temperature (°C)') g.set_titles(col_template="{col_name}") g.fig.suptitle('Max Temperature by Month for Each Year', y=1.03) plt.tight_layout() plt.show() # %% g = sns.FacetGrid(df_vis, col='year', col_wrap=4, height=3.5, aspect=1.2) g.map(sns.lineplot, 'month', 'temp_min', errorbar=None) g.set_axis_labels('Month', 'Min Temperature (°C)') g.set_titles(col_template="{col_name}") g.fig.suptitle('Min Temperature by Month for Each Year', y=1.03) plt.tight_layout() plt.show() # %% g = sns.FacetGrid(df_vis, col='year', col_wrap=4, height=3.5, aspect=1.2) g.map(sns.lineplot, 'month', 'precipitation', errorbar=None) g.set_axis_labels('Month', 'Precipitation (mm)') g.set_titles(col_template="{col_name}") g.fig.suptitle('Precipitation by Month for Each Year', y=1.03) plt.tight_layout() plt.show() # %% g = sns.FacetGrid(df_vis, col='year', col_wrap=4, height=3.5, aspect=1.2) g.map(sns.lineplot, 'month', 'wind', errorbar=None) g.set_axis_labels('Month', 'Wind Speed') g.set_titles(col_template="{col_name}") g.fig.suptitle('Wind Speed by Month for Each Year', y=1.03) plt.tight_layout() plt.show() # %% [markdown] # ### Weather Category Distribution # %% print_header("Weather Category Counts") weather_counts = df['weather'].value_counts() print(weather_counts) # %% plt.figure(figsize=(10, 6)) sns.countplot(data=df, x='weather', order=weather_counts.index, hue='weather', palette="viridis", legend=False) plt.title('Distribution of Weather Types') plt.xlabel('Weather Type') plt.ylabel('Frequency') plt.xticks(rotation=45) plt.show() # %% plt.figure(figsize=(10, 8)) plt.pie(weather_counts, labels=weather_counts.index, autopct='%1.1f%%', startangle=140, colors=sns.color_palette("viridis", len(weather_counts))) plt.title('Distribution of Weather Types (Pie Chart)') plt.axis('equal') plt.show() # %% [markdown] # ## Data Preprocessing for Classification # %% df_processed = df.copy() if 'date' in df_processed.columns: df_processed = df_processed.drop('date', axis=1) print_header("DataFrame columns before modeling") print(df_processed.columns.tolist()) df_processed.head() # %% le = LabelEncoder() df_processed['weather_encoded'] = le.fit_transform(df_processed['weather']) print_header("Label Encoding Mapping for 'weather'") for i, class_name in enumerate(le.classes_): print(f"{class_name} -> {i}") # %% joblib.dump(le, 'weather_label_encoder.joblib') print_header("Saved weather_label_encoder.joblib") df_processed.head() # %% [markdown] # ### Adding Lagged Time Series Features # %% df_for_lagged_processing = df.copy() # Start with the original df that includes 'date' le_lagged = LabelEncoder() # Use a separate encoder if necessary, or ensure le can be reused df_for_lagged_processing['weather_encoded'] = le_lagged.fit_transform(df_for_lagged_processing['weather']) # --- Feedback Implementation: Using split helper functions --- # Step 1: Create lag features. # The _create_lag_features function will sort by 'date' internally. # Default features_to_lag includes 'weather_encoded'. df_temp_lags = _create_lag_features( df_input=df_for_lagged_processing, lag_period=1 ) # Step 2: Create delta features using the DataFrame that now includes lags. # Default temp_features_for_delta are 'temp_max', 'temp_min'. df_with_lags_and_deltas_raw = _create_delta_features( df_input=df_temp_lags, lag_period=1 ) # Step 3: Drop rows with NaN values introduced by shifting and reset index df_with_lags = df_with_lags_and_deltas_raw.dropna().reset_index(drop=True) print_header("DataFrame with Lagged and Delta Features") print(df_with_lags.head()) df_processed_lagged = df_with_lags.drop(columns=['date', 'weather'], errors='ignore') print_header("df_processed_lagged head") print(df_processed_lagged.head()) # %% [markdown] # ## Feature Selection and Train-Test Split # %% original_features = ['temp_min', 'temp_max', 'precipitation', 'wind'] X_original = df_processed[original_features] # df_processed does not have 'date' y_original = df_processed['weather_encoded'] lagged_features_input_list = ['temp_min', 'temp_max', 'precipitation', 'wind', 'precipitation_lag1', 'temp_max_lag1', 'temp_min_lag1', 'wind_lag1', 'weather_encoded_lag1', 'delta_temp_max', 'delta_temp_min'] # Ensure all features in list are present in df_processed_lagged.columns # Filter to only include columns that actually exist in df_processed_lagged lagged_features_input_list = [col for col in lagged_features_input_list if col in df_processed_lagged.columns] X_lagged = df_processed_lagged[lagged_features_input_list] y_lagged = df_processed_lagged['weather_encoded'] feature_names_for_model = X_original.columns.tolist() joblib.dump(feature_names_for_model, 'classifier_feature_names.joblib') print(f"Saved classifier_feature_names.joblib with features: {feature_names_for_model}") # --- Feedback Implementation: Train/Test Split ratio --- # Changed test_size from 0.2 to 0.25 X_train_original, X_test_original, y_train_original, y_test_original = train_test_split( X_original, y_original, test_size=0.25, random_state=42, stratify=y_original ) print_header("Original Data Split Shapes") print(f"Original X_train shape: {X_train_original.shape}, y_train shape: {y_train_original.shape}") print(f"Original X_test shape: {X_test_original.shape}, y_test shape: {y_test_original.shape}") # Changed test_size from 0.2 to 0.25 X_train_lagged, X_test_lagged, y_train_lagged, y_test_lagged = train_test_split( X_lagged, y_lagged, test_size=0.25, random_state=42, stratify=y_lagged ) print_header("Lagged Data Split Shapes") print(f"Lagged X_train shape: {X_train_lagged.shape}, y_train shape: {y_train_lagged.shape}") print(f"Lagged X_test shape: {X_test_lagged.shape}, y_test shape: {y_test_lagged.shape}") # %% [markdown] # ## Naïve Model (Climate Prediction) # %% df_naive = df.copy() df_naive['month'] = df_naive['date'].dt.month monthly_most_frequent_weather = df_naive.groupby('month')['weather'].agg(lambda x: x.mode()[0] if not x.mode().empty else "Unknown") print_header("Most frequent weather type per month (Naïve Model)") print(monthly_most_frequent_weather) df_naive['predicted_weather_naive'] = df_naive['month'].map(monthly_most_frequent_weather) naive_accuracy = accuracy_score(df_naive['weather'], df_naive['predicted_weather_naive']) print(f"\nNaïve Model Accuracy (Predicting most frequent weather by month): {naive_accuracy:.4f}") print("This simple model predicts the 'climate' for each month, rather than specific 'weather'.") # %% [markdown] # ## Model Training and Evaluation # %% nb_model_original, nb_accuracy_original = train_and_evaluate_model( GaussianNB(), X_train_original, y_train_original, X_test_original, y_test_original, le.classes_, "Gaussian Naive Bayes (Original Features)" ) # Ensure le_lagged.classes_ is used for lagged model evaluation if its encoding differs. # If df['weather'].unique() is consistent, le.classes_ and le_lagged.classes_ will be the same. nb_model_lagged, nb_accuracy_lagged = train_and_evaluate_model( GaussianNB(), X_train_lagged, y_train_lagged, X_test_lagged, y_test_lagged, le_lagged.classes_, "Gaussian Naive Bayes (Lagged Features)" ) lr_model_original, lr_accuracy_original = train_and_evaluate_model( LogisticRegression(max_iter=1000, random_state=42), X_train_original, y_train_original, X_test_original, y_test_original, le.classes_, "Logistic Regression (Original Features)" ) svm_model_original, svm_accuracy_original = train_and_evaluate_model( SVC(random_state=42), X_train_original, y_train_original, X_test_original, y_test_original, le.classes_, "Support Vector Machine (Original Features)" ) # %% [markdown] # ## Ablation Study # %% features_to_ablate = [ ['temp_min', 'temp_max', 'precipitation', 'wind'], ['temp_min', 'temp_max', 'precipitation'], ['temp_min', 'temp_max', 'wind'], ['precipitation', 'wind'], ['temp_max', 'precipitation', 'wind'], ['temp_min', 'precipitation', 'wind'], ['temp_max'], ['wind'] ] ablation_results = {} print_header("--- Ablation Study (Gaussian Naive Bayes with Original Features) ---") for i, current_features in enumerate(features_to_ablate): print(f"\nTraining with features: {current_features}") X_ablation = df_processed[current_features] y_ablation = df_processed['weather_encoded'] # --- Feedback Implementation: Train/Test Split ratio --- # Changed test_size from 0.2 to 0.25 X_train_ab, X_test_ab, y_train_ab, y_test_ab = train_test_split( X_ablation, y_ablation, test_size=0.25, random_state=42, stratify=y_ablation ) model = GaussianNB() model.fit(X_train_ab, y_train_ab) y_pred_ab = model.predict(X_test_ab) accuracy_ab = accuracy_score(y_test_ab, y_pred_ab) ablation_results[tuple(current_features)] = accuracy_ab print(f"Accuracy: {accuracy_ab:.4f}") print_header("--- Ablation Study Summary ---") for features, acc in ablation_results.items(): print(f"Features: {features} -> Accuracy: {acc:.4f}") # %% [markdown] # ## Save the Model for Flask App # %% joblib.dump(nb_model_original, 'weather_prediction_model.joblib') print_header("Saved weather_prediction_model.joblib (Gaussian Naive Bayes with original features)") # %% [markdown] # ## Note on Further Model Exploration (Feedback Suggestion) # # The feedback included a suggestion to explore a Random Forest classifier. # Random Forests can be powerful and offer insights into feature importance. # # Key benefits: # - Often provide good performance with less hyperparameter tuning. # - Can handle a mix of numerical and categorical features (though scikit-learn requires encoding). # - Provide `feature_importances_` attribute, which can be used to understand which features the model found most predictive. This automates some aspects of an ablation study or feature selection. # - Generally robust to overfitting, especially with more trees. # # This would be a good next step to potentially improve predictive performance and gain further insights from the data. # You could train it similarly to the other models: # # ```python # from sklearn.ensemble import RandomForestClassifier # # # rf_model, rf_accuracy = train_and_evaluate_model( # # RandomForestClassifier(random_state=42, n_estimators=100), # # X_train_original, y_train_original, X_test_original, y_test_original, # # le.classes_, "Random Forest (Original Features)" # # ) # # print(f"Random Forest Feature Importances: {rf_model.feature_importances_}") # ``` # %% print_header("Notebook execution complete!") I just need to be sure if I have managed to implement the feedback accurately before transforming the code of Jupyter Notebook into a Flask app. Continue reading...
