3rd ICAI 2024

International Conference on Automotive Industry 2024

Mladá Boleslav, Czech Republic

2.2 Machine learning Machine learning algorithms extract patterns from data to enable predictions or decisions without explicit programming. The effectiveness of an ML initiative greatly hinges on the meticulous completion of various essential stages, commencing with data preprocessing and culminating in the deployment of the model within production settings. • Collection and Preparation of Data: Gathering and formatting data is paramount. Often, vast amounts are available, but they may be unstructured or noisy. Data cleaning and preprocessing are essential to transform it into a structured format suitable for analysis. • Feature Selection: Not all features in the data are relevant for learning. Therefore, it’s crucial to select a subset of the most important features while discarding irrelevant ones. This helps streamline the learning process and improve efficiency. • Choice of Algorithm: Different machine learning algorithms cater to different types of problems. It’s vital to choose the most suitable algorithm that aligns with the problem’s characteristics. Understanding the various algorithms and their applicability is key to achieving optimal results. • Selection of Models and Parameters: Many machine learning algorithms involve parameter tuning to achieve the best performance. Manual intervention is often required to determine the most appropriate values for these parameters, optimizing the model’s effectiveness. • Training: Once the algorithm and parameter values are set, the model undergoes training using a portion of the dataset designated as training data. This step allows the model to learn patterns and relationships within the data. • Performance Evaluation: Before deploying the model in real-world scenarios, it is essential to assess its performance. Testing the model against unseen data helps evaluate its efficacy, typically measured through various performance metrics such as accuracy, precision, and recall. This evaluation ensures that the model has learned effectively and can generalize well to new data. The routine outlined above, however, does not include an important step related to the definition of the machine learning issue class. A description of the classes for defining the machine learning issue class can be found in the work of m.a. (Peng et al., 2021). The following is a descriptive outline of the classes of machine learning issues and a summary in the form of a graphic presented in Figure 2. • Supervised Learning: In supervised learning, the algorithm is trained on a labelled dataset, meaning that each input is paired with the correct output. The goal is for the algorithm to learn the mapping between inputs and outputs, enabling it to make predictions or decisions when presented with new, unseen data. Common tasks in supervised learning include classification and regression.

82