Algorithm A3. DBN-based Fall Prediction Algorithm (DBN-FPA): Pseudocode of the Proposed Model [26]

Input: ADLs, Falls History—Datasets from public Repository. Output:Risk level classification—Fall Prediction Outcome Data extraction from the input parameters Feature extraction → Feed extracted data to the DBN-based prediction model. (a) Data Normalization (X) → input data with features (fall history, ADLs). Min-Max normalization to each feature in (X):                    $X^{\prime }={\displaystyle \frac{X-X_{min}}{X_{max}-X_{min}}}$ where $X_{min}$, and $X_{max}$ → minimum and maximum values of X (b) Encode Labels Y → target labels using label encoding (c)Split normalized data → X’ encoded labels → Y data into training and test sets (d) Pre-train with RBMs Train RBM_1: - components=256, learning_rate=0.01, _iter=10 Train RBM_2: - components=200, learning_rate=0.01, _iter=10 Train RBM_3: - components=150, learning_rate=0.01, _iter=10 (e)Fine-tuning → Neural Network + Adam Optimizer (f) Initialize the weights of the neural network with the weights from the pre-trained RBMs:   - Model_layers [0] → set_weights → RBM_1 - Model_layers [1] → set_weights → RBM_2 - Model_layers [2] → set_weights → RBM_3 3.Train using DBN- Risk Calculation 4.Cross-validation using testing dataset → Determine the accuracy, specificity, and sensitivity of the model. 5.Result Evaluation → validate the findings.