Note:Given the substantial volume of data, even with the utilization of TPUs, it remains challenging to allocate sufficient memory.","metadata":{}},{"cell_type":"code","source":"def split_dataset(dataset, test_ratio=0.30):\n test_indices = np.random.rand(len(dataset)) < test_ratio\n return dataset[~test_indices], dataset[test_indices]\n\n\ntrain_ds_pd, valid_ds_pd = split_dataset(train)\nprint(\"{} examples in training, {} examples in testing.\".format(\n len(train_ds_pd), len(valid_ds_pd)))","metadata":{"execution":{"iopub.status.busy":"2023-09-16T15:10:39.589435Z","iopub.execute_input":"2023-09-16T15:10:39.589871Z","iopub.status.idle":"2023-09-16T15:10:57.134853Z","shell.execute_reply.started":"2023-09-16T15:10:39.589831Z","shell.execute_reply":"2023-09-16T15:10:57.131364Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"label = 'event'\ntrain_ds = tfdf.keras.pd_dataframe_to_tf_dataset(train_ds_pd, label=label, task = tfdf.keras.Task.CLASSIFICATION)\nvalid_ds = tfdf.keras.pd_dataframe_to_tf_dataset(valid_ds_pd, label=label, task = tfdf.keras.Task.CLASSIFICATION)","metadata":{"execution":{"iopub.status.busy":"2023-09-16T15:10:57.137459Z","iopub.execute_input":"2023-09-16T15:10:57.137842Z","iopub.status.idle":"2023-09-16T15:12:51.412726Z","shell.execute_reply.started":"2023-09-16T15:10:57.137810Z","shell.execute_reply":"2023-09-16T15:12:51.410602Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"tfdf.keras.get_all_models()","metadata":{"execution":{"iopub.status.busy":"2023-09-16T15:12:51.415529Z","iopub.execute_input":"2023-09-16T15:12:51.416114Z","iopub.status.idle":"2023-09-16T15:12:51.428259Z","shell.execute_reply.started":"2023-09-16T15:12:51.416063Z","shell.execute_reply":"2023-09-16T15:12:51.427049Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"rf = tfdf.keras.RandomForestModel()\nrf.compile(metrics=[\"precision\"])","metadata":{"execution":{"iopub.status.busy":"2023-09-16T15:12:51.429674Z","iopub.execute_input":"2023-09-16T15:12:51.430073Z","iopub.status.idle":"2023-09-16T15:12:51.592095Z","shell.execute_reply.started":"2023-09-16T15:12:51.430043Z","shell.execute_reply":"2023-09-16T15:12:51.590925Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"rf.fit(x=train_ds)","metadata":{"execution":{"iopub.status.busy":"2023-09-16T15:12:51.593912Z","iopub.execute_input":"2023-09-16T15:12:51.594634Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"tfdf.model_plotter.plot_model_in_colab(rf, tree_idx=0, max_depth=1)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_ds = tfdf.keras.pd_dataframe_to_tf_dataset(\n test,\n task = tfdf.keras.Task.CLASSIFICATION)\n\n\ntest_ds = tfdf.keras.pd_dataframe_to_tf_dataset(\n test,\n task = tfdf.keras.Task.CLASSIFICATION)\n\npreds = rf.predict(test_ds)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"importance_metrics = [\"SUM_SCORE\", \"INV_MEAN_MIN_DEPTH\", \"NUM_NODES\", \"NUM_AS_ROOT\"]\ninspector = rf.make_inspector()\ninspector.evaluation()\nimportances = inspector.variable_importances()\n# Define the number of top variables to visualize\ntop_variables_count = 4\n\n# Create a 2x2 grid of subplots with improved styling\nfig, axes = plt.subplots(2, 2, figsize=(14, 10))\nfig.suptitle(\"Variable Importances for Different Metrics\", fontsize=16)\n\n# Iterate through importance metrics and plot in subplots\nfor i, importance_metric in enumerate(importance_metrics):\n # Extract the feature names and importance values for the current metric\n variable_importances = importances[importance_metric]\n \n # Sort by importance values and select the top 12 variables\n variable_importances.sort(key=lambda x: x[1], reverse=True)\n top_variable_importances = variable_importances[:top_variables_count]\n \n feature_names = [vi[0].name for vi in top_variable_importances]\n feature_importances = [vi[1] for vi in top_variable_importances]\n \n # Determine the row and column for the current subplot\n row = i // 2\n col = i % 2\n\n # Create a bar chart for the current metric in the specified subplot with improved styling\n axes[row, col].barh(feature_names, feature_importances)\n axes[row, col].invert_yaxis()\n axes[row, col].set_xlabel(\"Importance Score\")\n axes[row, col].set_title(f\"Top {top_variables_count} Variables for {importance_metric}\", fontsize=12)\n\n # Add labels to the bars\n for idx, importance in enumerate(feature_importances):\n axes[row, col].text(importance, idx, f\"{importance:.4f}\", va='center', fontsize=10, color='black')\n\nplt.tight_layout(pad=3.0)\nplt.show()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"try:\n result_df = test.copy() # Copy the test DataFrame\n result_df['Score'] = preds\n result_df.to_parquet(\"Score.parquet\",index=False)\nexcept Exception as e:\n print(\"Error saving\")","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample_submission_df = pd.read_csv('/kaggle/input/child-mind-institute-detect-sleep-states/sample_submission.csv')\nsample_submission_df['Score'] = rf.predict(test_ds)\nsample_submission_df.to_csv('/kaggle/working/submission.csv', index=False)\nsample_submission_df.head()","metadata":{"_kg_hide-input":true,"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":" \nIf you find value in this notebook, I would greatly appreciate it if you could consider giving it a like, following me for more updates, and sharing it with others who might also benefit from it.\n\nYour support means a lot to me and helps me continue sharing insights and knowledge within the Kaggle community.\n\n \nThank you!\n
","metadata":{}}]}