mininato/EmotionClassificationPipeline

1# Configuration file for the pipeline2 3config = {4    # Paths for Import Data5    "accel_path": "/Users/anhducduong/Documents/GitHub/EmotionRecognitionPipeline/EmotionRecognitionPipeline/AccelerometerMeasurements_backup.csv",  # Path to the accelerometer data6    "reports_path": "/Users/anhducduong/Documents/GitHub/EmotionRecognitionPipeline/EmotionRecognitionPipeline/UserTestingSelfReports.csv",  # Path to the self-reports data7    #"combined_data_path": "Path or Name of File of Combined Data File",  # Path to the combined data8    #"features_data_path": "Path or Name of File of Features Data File",  # Path to the features data9    #"model_path": "Path or Name of Trained Model File",  # Path to the trained model10 11    # Label Configuration12    "label_columns": ["valence", "arousal"],  # Here you should input the emotion-labels that you are using13    "target_label": "arousal",  # This is the target label that you want to predict (Only one label can be selected)14 15    # Configuration for combined data16    "time_window": 3,  # Minutes before and after the self-report17 18    # Configuration for feature extraction19    "window_length": 60,  # Window length in seconds / 6020    "window_step_size": 20,  # Step size in seconds / 10%-50% of window_length / 2021    "data_frequency": 25,  # Data frequency in Hz22    "selected_domains": None,  # Default: Every domain / 'time_domain', 'spatial', 'frequency', 'statistical', 'wavelet' / multiple domains: ["time_domain", "frequency"] / order is not important23    "include_magnitude": True,  # Include magnitude-based features or not24 25    #Configuration for Low-pass filter26    "cutoff_frequency": 10,  # Cut-off frequency for the low-pass filter27    "order": 4,  # Order of the filter28 29    # Configuration for Scaling30    "scaler_type": "standard",  # Possible Scaler: 'standard' or 'minmax'31 32    # Configuration for PCA33    "apply_pca": False,  # Apply PCA or not34    "pca_variance": 0.95,  # PCA variance threshold35 36    # Configuration for model training37    "classifier": "xgboost",  # Default classifier ('xgboost', 'svm', 'randomforest')38 39    # Configuration for hyperparameter tuning40    "n_splits": 5, # Number of splits for cross-validation41    "n_iter": 30,   # Number of iterations for hyperparameter tuning42    "n_jobs": -1,   # Number of jobs for parallel processing43    "n_points": 1,  # Number of points to sample in the hyperparameter space44 45    # If users want to define custom param_space, they can specify it here46    "param_space": {47        "learning_rate": (0.05, 0.2), 48        "n_estimators": (200, 800),49        "max_depth": (4, 8),50        "min_child_weight": (1, 5),51        "subsample": (0.6, 0.9),52        "colsample_bytree": (0.6, 0.9),53        "gamma": (0, 5),54        "reg_alpha": (0, 5),55        "reg_lambda": (0, 5)56    },  # Set to {None} to use default inside the TrainModel class57}