RFC

model/utils/Precision.ipynb

@@ -1998,7 +1998,7 @@
     "output_cleared": false,
     "deepnote_cell_type": "code"
    },
-   "source": "model_type: \"ANN\" or \"CNN\" or \"LSTM\" or \"RFC\" = \"CNN\"\nids = [1]\nPRINT_CLASSIFICATION_MATRIX = False\n\nfor run_i in ids:\n    if model_type != \"RFC\":\n        model = load_model(f\"../models/{model_type}_{AMOUNT_OF_SENSORS}_sensor{'er' if AMOUNT_OF_SENSORS > 1 else ''}_{run_i}.h5\")\n    else:\n        model = load(f\"../models/RFC_model_{AMOUNT_OF_SENSORS}.joblib\")\n    print(\"Classification accuracy:\")\n    classification_dict = dict()\n    accuracy_list = list() \n\n    for key in x_test_dict:\n        x_test_numpy = x_test_dict[key].drop([' TimeStamp (s)', 'Pose'], axis=1).values\n        if model_type == \"RFC\": \n            x_test_numpy = x_test_dict[key].drop([' TimeStamp (s)', 'Pose'], axis=1)\n        if model_type == \"CNN\":\n            x_test_numpy = x_test_numpy.reshape(x_test_numpy.shape[0],x_test_numpy.shape[1], 1)\n        classify = model.predict(x_test_numpy)\n        classifications = [i.argmax() for i in classify]\n        if model_type == 'RFC': classifications = classify\n        annotated_positions = y_test_dict[key].to_numpy()\n        correct_classifications = (classifications == annotated_positions).sum()\n        accuracy_list.append(round(correct_classifications/len(classifications)*100,2))\n        print(f\"{key}: {accuracy_list[-1]}%\")\n        classification_dict[key] = classifications\n    print(f\"Average accuracy: {round(sum(accuracy_list)/len(accuracy_list),2)}%\")\n\n    if PRINT_CLASSIFICATION_MATRIX:\n        width=16\n        height=4*len(x_test_dict)\n        fig, axes = plt.subplots(len(x_test_dict), 4, figsize=(width, height))\n        for i, key in enumerate(x_test_dict):\n            cols = [f\"{key}: Model classifications\", f\"{key}: Model heatmap\", f\"{key}: Annotated classifications\", f\"{key}: Annotated heatmap\"]\n            df_predict = pd.DataFrame({' Timestamp (s)': x_test_dict[key][' TimeStamp (s)'],'Pose':classification_dict[key]})\n            sns.lineplot(ax=axes[i, 0], data=df_predict,x=' Timestamp (s)',y='Pose')\n            sns.heatmap(ax=axes[i, 1], data=confusion_matrix(y_test_dict[key], classification_dict[key]), cmap=\"YlGnBu\", annot=True, fmt=\"d\")\n            sns.lineplot(ax=axes[i, 2], data=x_test_dict[key], x=\" TimeStamp (s)\", y='Pose')\n            sns.heatmap(ax=axes[i, 3], data=confusion_matrix(y_test_dict[key], y_test_dict[key].to_numpy()), cmap=\"YlGnBu\", annot=True, fmt=\"d\")\n            for ax, col in zip(axes[i], cols): ax.set_title(col)\n        fig.tight_layout()\n        plt.show()\n        fig.savefig(f\"{model_type}_plot.png\")",
+   "source": "model_type: \"ANN\" or \"CNN\" or \"LSTM\" or \"RFC\" = \"CNN\"\nids = [1]\nPRINT_CLASSIFICATION_MATRIX = False\n\nfor run_i in ids:\n    if model_type != \"RFC\":\n        model = load_model(f\"../models/{model_type}_{AMOUNT_OF_SENSORS}_sensor{'er' if AMOUNT_OF_SENSORS > 1 else ''}_{run_i}.h5\")\n    else:\n        model = load(f\"../models/RFC_model_{AMOUNT_OF_SENSORS}.joblib\")\n    print(\"Classification accuracy:\")\n    classification_dict = dict()\n    accuracy_list = list() \n\n    for key in x_test_dict:\n        x_test_numpy = x_test_dict[key].drop([' TimeStamp (s)', 'Pose'], axis=1).values\n        if model_type == \"RFC\": \n            x_test_numpy = x_test_dict[key].drop([' TimeStamp (s)', 'Pose'], axis=1)\n        if model_type == \"CNN\":\n            x_test_numpy = x_test_numpy.reshape(x_test_numpy.shape[0],x_test_numpy.shape[1], 1)\n        classify = model.predict(x_test_numpy)\n        classifications = [i.argmax() for i in classify]\n        if model_type == 'RFC': classifications = classify\n        annotated_positions = y_test_dict[key].to_numpy()\n        correct_classifications = (classifications == annotated_positions).sum()\n        accuracy_list.append(round(correct_classifications/len(classifications)*100,2))\n        print(f\"{key}: {accuracy_list[-1]}%\")\n        classification_dict[key] = classifications\n    print(f\"Average accuracy: {round(sum(accuracy_list)/len(accuracy_list),2)}%\")\n    \n\n    if PRINT_CLASSIFICATION_MATRIX:\n        width=16\n        height=4*len(x_test_dict)\n        fig, axes = plt.subplots(len(x_test_dict), 4, figsize=(width, height))\n        for i, key in enumerate(x_test_dict):\n            cols = [f\"{key}: Model classifications\", f\"{key}: Model heatmap\", f\"{key}: Annotated classifications\", f\"{key}: Annotated heatmap\"]\n            df_predict = pd.DataFrame({' Timestamp (s)': x_test_dict[key][' TimeStamp (s)'],'Pose':classification_dict[key]})\n            sns.lineplot(ax=axes[i, 0], data=df_predict,x=' Timestamp (s)',y='Pose')\n            sns.heatmap(ax=axes[i, 1], data=confusion_matrix(y_test_dict[key], classification_dict[key]), cmap=\"YlGnBu\", annot=True, fmt=\"d\")\n            sns.lineplot(ax=axes[i, 2], data=x_test_dict[key], x=\" TimeStamp (s)\", y='Pose')\n            sns.heatmap(ax=axes[i, 3], data=confusion_matrix(y_test_dict[key], y_test_dict[key].to_numpy()), cmap=\"YlGnBu\", annot=True, fmt=\"d\")\n            for ax, col in zip(axes[i], cols): ax.set_title(col)\n        fig.tight_layout()\n        plt.show()\n        fig.savefig(f\"{model_type}_plot.png\")",
    "execution_count": 5,
    "outputs": [
     {