This project implements a custom neural network from scratch using a sigmoid activation function and a linear hypothesis. The model was trained and tested on the Heart Disease Dataset, achieving an accuracy of 65%.
- Custom Neural Network: Built without relying on high-level libraries, providing a deeper understanding of neural network structure and training.
- Activation Function: Utilizes the sigmoid function for neuron activations, ideal for binary classification tasks.
- Dataset: Tested on the Heart Disease Dataset, offering insight into how neural networks can predict the likelihood of heart disease.
- Activation Function: Sigmoid function
- Formula:
- Formula:
- Hypothesis: Linear hypothesis model, where:
- Output = Activation(Weights * Input + Biases)
- Training Parameters:
- Epochs: 1000
- Learning Rate: 0.01
The model achieved 65% accuracy on the test set, showing promising results in predicting heart disease.
- Experiment with more complex architectures to potentially increase accuracy.
- Adjust hyperparameters, such as learning rate and epochs, for performance optimization.
- Data preprocessing techniques and normalization could also enhance model performance.