Performance Metrics

Improve your Bayesian network's performance and accuracy by learning key data-based performance metrics.

We'll cover the following

Welcome to this section on tuning data-based hyperparameters of a Bayesian network. In this lesson, we will explore various techniques and approaches to improve the performance and accuracy of our Bayesian network model. The topics we will cover in this section include:

  • Features optimization: We'll introduce the key data-based performance metrics for Bayesian models and demonstrate how to evaluate the model's performance based on these metrics, particularly using the Receiver Operating Characteristic (ROC) curve.

  • Discretize data: We'll explore data preprocessing and preparation techniques such as discretization to ensure accurate and suitable data for training and inference in BN models.

  • AIC and BIC curve: We'll learn to interpret AIC and BIC curves for comparing and evaluating BN structures, helping us choose the best model for our data.

  • Parameter learning algorithms: We'll cover various algorithms like Maximum Likelihood Estimation (MLE) and Bayesian Estimation for estimating conditional probability distributions in BNs.

  • Structure learning algorithms: We'll cover constraint-based, score-based, and hybrid methods for learning BN structures, discussing the advantages and limitations of each approach.

  • Inference algorithms: We'll explore inference algorithms like Exact Inference, Approximate Inference, and Markov Chain Monte Carlo (MCMC), discussing the trade-offs between efficiency and accuracy to choose the best algorithm for our BN model.

Performance metrics

Let’s discuss the performance metrics of BN learning with a focus on data-based hyperparameters. While accuracy is an important measure, it is crucial to consider additional performance metrics for a comprehensive evaluation of the model's performance. By understanding and combining these metrics, we will be better equipped to assess and optimize our BN model's performance effectively.

To start with the following definitions:

  • True positive (TP): The number of instances where the model correctly predicts a positive outcome.

  • True negative (TN): The number of instances where the model correctly predicts a negative outcome.

  • False positive (FP): The number of instances where the model incorrectly predicts a positive outcome when it is actually negative.

  • False negative (FN): The number of instances where the model incorrectly predicts a negative outcome when it is actually positive.

  • Total predictions (TO): The total number of predictions made by the model, which includes true positives, true negatives, false positives, and false negatives.

Using these definitions we can better understand the performance metrics we are going to use:

Get hands-on with 1300+ tech skills courses.