GridDB, a high-performance NoSQL database, is particularly suited for managing complex and dynamic datasets. Its efficient in-memory processing and flexible data storage capabilities make it an ideal choice for real-time applications.<\/p>\n
Note:<\/strong> The codes for this tutorial are in my GridDB Blogs GitHub repository<\/a>.<\/p>\n You need to install the following libraries to run codes in this article.<\/p>\n To install these libraries, follow the installation instructions on GridDB Python Package Index (Pypi).<\/a><\/p>\n The code is executed in Google Colab<\/a>, so you do not need to install other libraries.<\/p>\n Run the following script to import the required libraries into your Python application.<\/p>\n We will begin by inserting the stress detection dataset from Kaggle into GridDB. In a later section, we will retrieve data from the GridDB and train our machine-learning algorithms for user stress prediction.<\/p>\n You can download the stress detection dataset from Kaggle<\/a> and import it into your Python application.<\/p>\n Output:<\/strong><\/p>\n The dataset consists of 3000 records belonging to 100 users. For each user, 30 days of data are recorded for various attributes such as openness, sleep duration, screen time, mobility distance, and number of calls.<\/p>\n The The following script displays various statistics for the Output:<\/strong><\/p>\n Next, we will insert the user stress dataset into GridDB.<\/p>\n You need to connect to a GridDB instance before inserting data into the GridDB.<\/p>\n To do so, you must create a GridDB factory instance using the Next, you have to call the To test whether the connection is successful, we retrieve a dummy container, Output:<\/strong><\/p>\n You should see the above message if the connection is successful.<\/p>\n GridDB stores data in containers, which are specialized data structures for efficient data structure.<\/p>\n The following script creates a GridDB container to store our stress detection dataset.<\/p>\n We first remove any existing container with the name Next, we replace empty spaces in column names with an underscore since GridDB does not expect column names to have spaces.<\/p>\n We will then map Pandas DataFrame data type to GridDB data types and create a column info list containing column names and corresponding GridDB data types,<\/p>\n Next, we create a container info object and pass it the container name, the column info list, and the container type, which is Finally, we call the grid store’s Output:<\/strong><\/p>\n The above message shows that the container creation is successful.<\/p>\n We can now store data in the container we created. The following script inserts our stress detection dataset into the Output:<\/strong><\/p>\n The above output shows that data insertion is successful.<\/p>\n In this section, we will retrieve the stress detection dataset from the The following script defines the Next, we create a Output:<\/strong><\/p>\n The above output shows the stress detection dataset we retrieved from the GridDB container.<\/p>\n We will first try to predict the The following script divides the dataset into features and labels, splits it into training and test sets, and normalizes it using the standard scaling approach.<\/p>\n Next, we create an object of the Finally, we evaluate the model performance by prediting Output:<\/strong><\/p>\n The output shows that, on average, our model’s predicted Next, we will develop a deep neural network for stress detection prediction.<\/p>\n We will use the TensorFlow<\/a> and Keras<\/a> libraries to create a sequential deep learning model with three dense layers. We will also add batch normalization and dropout to reduce model overfitting.<\/p>\n We will also use an adaptive learning rate so the gradient doesn’t overshoot while training.<\/p>\n Finally, we compile the model using the mean squared error loss and mean absolute error metric. We use this loss and metric since we are dealing with a regression problem.<\/p>\n Output:<\/strong><\/p>\n The above output shows the model summary.<\/p>\n Next, we train the model using the Output:<\/strong><\/p>\n We load the best model to evaluate the performance and use it to make predictions on the test set.<\/p>\nPrerequisites<\/h2>\n
\n
\nimport pandas as pd\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.metrics import mean_absolute_error, mean_squared_error\nfrom sklearn.ensemble import RandomForestRegressor\n\nimport tensorflow as tf\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Dense, Dropout, BatchNormalization\nfrom tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\nfrom tensorflow.keras.optimizers import Adam\nfrom tensorflow.keras.models import load_model\n\nimport griddb_python as griddb<\/code><\/pre>\n<\/div>\nInserting Stress Detection Dataset into GridDB<\/h2>\n
Downloading and Importing the Stress Detection Dataset from Kaggle<\/h3>\n
\n# Dataset download link\n# https:\/\/www.kaggle.com\/datasets\/swadeshi\/stress-detection-dataset?resource=download\n\ndataset = pd.read_csv(\"stress_detection.csv\")\nprint(f\"The dataset consists of {dataset.shape[0]} rows and {dataset.shape[1]} columns\")\ndataset.head()<\/code><\/pre>\n<\/div>\n![\"\"](\"https:\/\/griddb.net\/wp-content\/uploads\/2025\/06\/img1-user-stress-dataset.webp\")
<\/a><\/p>\nPSS_score<\/code> column contains the perceived stress score, which ranges from 10 to 40. A higher score corresponds to a higher stress level.<\/p>\nPSS_score<\/code> column.<\/p>\n\ndataset[\"PSS_score\"].describe()<\/code><\/pre>\n<\/div>\n\ncount 3000.000000\nmean 24.701000\nstd 8.615781\nmin 10.000000\n25% 17.000000\n50% 25.000000\n75% 32.000000\nmax 39.000000\nName: PSS_score, dtype: float64\n<\/code><\/pre>\n<\/div>\nConnect to GridDB<\/h3>\n
griddb.StoreFactory.get_instance()<\/code> method.<\/p>\nget_store<\/code> method on the factory instance and pass the database host URL, cluster name, and user name and password. The get_store()<\/code> method returns a grid store<\/code> object that you can use to create containers in GridDB.<\/p>\ncontainer1<\/code>, as shown in the script below.<\/p>\n\n# GridDB connection details\nDB_HOST = \"127.0.0.1:10001\"\nDB_CLUSTER = \"myCluster\"\nDB_USER = \"admin\"\nDB_PASS = \"admin\"\n\n# creating a connection\n\nfactory = griddb.StoreFactory.get_instance()\n\ntry:\n gridstore = factory.get_store(\n notification_member = DB_HOST,\n cluster_name = DB_CLUSTER,\n username = DB_USER,\n password = DB_PASS\n )\n\n container1 = gridstore.get_container(\"container1\")\n if container1 == None:\n print(\"Container does not exist\")\n print(\"Successfully connected to GridDB\")\n\nexcept griddb.GSException as e:\n for i in range(e.get_error_stack_size()):\n print(\"[\", i, \"]\")\n print(e.get_error_code(i))\n print(e.get_location(i))\n print(e.get_message(i))<\/code><\/pre>\n<\/div>\n\nContainer does not exist\nSuccessfully connected to GridDB<\/code><\/pre>\n<\/div>\nCreate Container for User Stress Data in GridDB<\/h3>\n
user_stress_data<\/code> as we will use this name to create a new container.<\/p>\nCOLLECTION<\/code> for tabular data.<\/p>\nput_container<\/code> method and pass the container info object we created to it as a parameter.<\/p>\n\n# drop container if already exists\ngridstore.drop_container(\"user_stress_data\")\n\n# Clean column names to remove spaces or forbidden characters in the GridDB container\ndataset.columns = [col.strip().replace(\" \", \"_\") for col in dataset.columns]\n\n\n# Mapping from pandas data types to GridDB data types\ntype_mapping = {\n 'float64': griddb.Type.DOUBLE,\n 'int64': griddb.Type.INTEGER,\n 'object': griddb.Type.STRING,\n 'category': griddb.Type.STRING # Map category to STRING for GridDB\n}\n\n\n# Generate column_info dynamically\ncolumn_info = [[col, type_mapping[str(dtype)]] for col, dtype in dataset.dtypes.items()]\n\n# Define the container info\ncontainer_name = \"user_stress_data\"\ncontainer_info = griddb.ContainerInfo(\n container_name, column_info, griddb.ContainerType.COLLECTION, row_key=True\n)\n\n# Connecting to GridDB and creating the container\ntry:\n gridstore.put_container(container_info)\n container = gridstore.get_container(container_name)\n if container is None:\n print(f\"Failed to create container: {container_name}\")\n else:\n print(f\"Successfully created container: {container_name}\")\n\nexcept griddb.GSException as e:\n print(f\"Error creating container {container_name}:\")\n for i in range(e.get_error_stack_size()):\n print(f\"[{i}] Error code: {e.get_error_code(i)}, Message: {e.get_message(i)}\")<\/code><\/pre>\n<\/div>\n\nSuccessfully created container: user_stress_data<\/code><\/pre>\n<\/div>\nInsert User Stress Data into GridDB<\/h3>\n
\nTo do so, we iterate through the rows in our dataset, convert the column data into a GridDB data type, and store each row in the container using the put()<\/code> method.<\/p>\nuser_stress_data<\/code> container we created.<\/p>\n\ntry:\n for _, row in dataset.iterrows():\n # Prepare each row's data in the exact order as defined in `column_info`\n row_data = [\n int(row[col]) if dtype == griddb.Type.INTEGER else\n float(row[col]) if dtype == griddb.Type.DOUBLE else\n str(row[col])\n for col, dtype in column_info\n ]\n # Insert the row data into the container\n container.put(row_data)\n\n print(f\"Successfully inserted {len(dataset)} rows of data into {container_name}\")\n\nexcept griddb.GSException as e:\n print(f\"Error inserting data into container {container_name}:\")\n for i in range(e.get_error_stack_size()):\n print(f\"[{i}] Error code: {e.get_error_code(i)}, Message: {e.get_message(i)}\")<\/code><\/pre>\n<\/div>\n\nSuccessfully inserted 3000 rows of data into user_stress_data<\/code><\/pre>\n<\/div>\nStress Detection Using Machine and Deep Learning<\/h2>\n
user_stress_data<\/code> GridDB container we created earlier. Subsequently, we will train machine learning and deep learning models for stress prediction.<\/p>\nRetrieving Data from GridDB<\/h3>\n
retrieve_data_from_griddb()<\/code> function that accepts the container name as a parameter and calls the get_container()<\/code> function on the grid store to retrieve the data container.<\/p>\nSELECT<\/code> query object and call its fetch()<\/code> method to retrieve all records from the user_stress_data<\/code> container. Finally, we call the fetch_rows()<\/code> function to convert the records into a Pandas DataFrame.<\/p>\n\ndef retrieve_data_from_griddb(container_name):\n\n try:\n data_container = gridstore.get_container(container_name)\n\n # Query all data from the container\n query = data_container.query(\"select *\")\n rs = query.fetch()\n\n data = rs.fetch_rows()\n return data\n\n except griddb.GSException as e:\n print(f\"Error retrieving data from GridDB: {e.get_message()}\")\n return None\n\n\nstress_dataset = retrieve_data_from_griddb(\"user_stress_data\")\nstress_dataset.head()\n<\/code><\/pre>\n<\/div>\n![\"\"](\"https:\/\/griddb.net\/wp-content\/uploads\/2025\/06\/img2-user-stress-dataset-from-griddb.webp\")
<\/a><\/p>\nPredicting User Stress with Machine Learning<\/h3>\n
PSS_score<\/code> using a traditional machine learning algorithm such as Random Forest Regressor<\/a>.<\/p>\n\nX = stress_dataset.drop(columns=['PSS_score'])\ny = stress_dataset['PSS_score']\n\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)\n\nscaler = StandardScaler()\nX_train = scaler.fit_transform(X_train)\nX_test = scaler.transform(X_test)<\/code><\/pre>\n<\/div>\nRandomForestRegressor<\/code> class from he Scikit learn<\/a> library and pass the training and test sets to the fit()<\/code> method.<\/p>\n\nrf_model = RandomForestRegressor(random_state=42, n_estimators=1000)\nrf_model.fit(X_train, y_train)<\/code><\/pre>\n<\/div>\nPSS_score<\/code> on the test set.<\/p>\n\nrf_predictions = rf_model.predict(X_test)\n\n# Evaluate the regression model\nmae = mean_absolute_error(y_test, rf_predictions)\n\nprint(f\"Mean Absolute Error: {mae:.4f}\")<\/code><\/pre>\n<\/div>\n\nMean Absolute Error: 7.8973<\/code><\/pre>\n<\/div>\nPSS_score<\/code> is off by 7.8 points. This is not so bad, but it is not very good either.<\/p>\nPredicting User Stress with Deep Learning<\/h2>\n
\nmodel = Sequential([\n Dense(128, input_dim=X_train.shape[1], activation='relu'),\n BatchNormalization(),\n Dropout(0.2),\n Dense(64, activation='relu'),\n BatchNormalization(),\n Dropout(0.2),\n Dense(32, activation='relu'),\n BatchNormalization(),\n Dropout(0.1),\n Dense(1) \n])\n\n# Adaptive learning rate scheduler with exponential decay\ninitial_learning_rate = 0.001\nlr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(\n initial_learning_rate=initial_learning_rate,\n decay_steps=10000,\n decay_rate=0.9\n)\n\n# Compile the model with Adam optimizer and a regression loss\nmodel.compile(optimizer=Adam(learning_rate=lr_schedule),\n loss='mean_squared_error',\n metrics=['mean_absolute_error'])\n\n# Summary of the model\nmodel.summary()\n<\/code><\/pre>\n<\/div>\n![\"\"](\"https:\/\/griddb.net\/wp-content\/uploads\/2025\/06\/img3-deep-learning-stress-detection-model-summary.webp\")
<\/a><\/p>\nfit()<\/code> method. We use an early stopping approach that stops model training if the loss doesn’t decrease for 100 consecutive epochs. Finally, we save the best model at the end of model training.<\/p>\n\n# Define callbacks for training\nearly_stopping = EarlyStopping(monitor='val_loss', patience=100, restore_best_weights=True)\nmodel_checkpoint = ModelCheckpoint('best_model.keras', monitor='val_loss', save_best_only=True)\n\n# Train the model with the callbacks\nhistory = model.fit(\n X_train, y_train,\n epochs=1000,\n batch_size=4,\n validation_data=(X_test, y_test),\n callbacks=[early_stopping, model_checkpoint],\n verbose=1\n)<\/code><\/pre>\n<\/div>\n![\"\"](\"https:\/\/griddb.net\/wp-content\/uploads\/2025\/06\/img4-deep-learning-model-training-results.webp\")
<\/a><\/p>\n\n\n# Load the best model\nbest_model = load_model('best_model.keras')\n\n# Make predictions on the test set\ny_pred = best_model.predict(X_test)\n\n# Calculate Mean Absolute Error\nmae = mean_absolute_error(y_test, y_pred)\nprint(f\"Mean Absolute Error: {mae:.4f}\")\n\n# Plot training history to show MAE over epochs\nplt.plot(history.history['mean_absolute_error'], label='Train MAE')\nplt.plot(history.history['val_mean_absolute_error'], label='Validation MAE')\nplt.title('Mean Absolute Error over Epochs')\nplt.xlabel('Epochs')\nplt.ylabel('MAE')\nplt.legend()\nplt.show()<\/code><\/pre>\n<\/div>\n
![\"\"](\"https:\/\/griddb.net\/wp-content\/uploads\/2025\/06\/img5-deep-learning-predictions.webp\")
<\/a><\/p>\n