Knowledge Extraction and Machine Learning

The recommended Basic Knowledge (BC) (in Portuguese, “Conhecimento Base”, designated by the acronym CB) is as follows:

• CB1. Knowledge of descriptive and inferential statistics, namely measures of central tendency and dispersion, probability distributions, correlation, estimation, and hypothesis testing.
• CB2. Basic linear algebra, including vectors, matrices, and solving linear systems, sufficient to understand data representation and transformation.
• CB3. Data manipulation and preparation, including importing data in common formats (CSV, Excel, or equivalents), data cleaning, handling missing values, transforming variables, and using data tables.
• CB4. Data visualization and exploratory analysis, including the construction and interpretation of tables, histograms, scatter plots, bar charts, and box plots.
• CB5. Technical reading in English, sufficient to understand software documentation, introductory scientific articles, and bibliography in the field.

The classes will be taught in a theoretical-practical format, and the teaching methodology will include various pedagogical methods, namely the expository method, the demonstrative method, and project-based learning.

The Teaching Methods (in Portuguese, “Métodos de Ensino” or the acronym ME) will be as follows:

• ME1. Expository method: The expository method will be used to present the concepts and main contents of the curricular unit. The teacher organizes and orally presents the content, structuring the reasoning and the result to be obtained. This presentation will be supported by slides, which will later be made available to students. This presentation will be complemented with some provided references.
• ME2. Demonstrative method: The demonstrative method will be used to exemplify some applications of concepts, namely the application of the various techniques addressed for each task of data mining and machine learning. Based on several practical worksheets provided, the teacher shares their know-how and demonstrates and assists students in their execution, so that they successfully perform what is requested, sometimes on paper, other times on a computer, using a specific tool for this purpose.
• ME3. Project-based learning: The project-based learning method will be used to build knowledge through a long and continuous study process, the purpose of which is to address a challenge/problem whose objective is the development of a data mining and machine learning project, using data from an organization (private or public), open data, or creating data through surveys.

This course unit aims to teach the fundamentals of knowledge extraction (KE) and machine learning (ML), helping to create and interpret models that, by incorporating diverse and vast data, allow organizations to detect standard behaviours and future trends, make proactive decisions, and identify opportunities or risks.

The main Learning Objectives (in Portuguese “Objetivos de Aprendizagem”, designated by the acronym OA) are:

• OA1. To explore the concepts, application domains, and opportunities of KE and ML
  
• OA2. To present and apply in practice the main knowledge associated with the most common tasks and techniques
  
• OA3. To develop a KE/ML project according to an appropriate methodology

The main Competencies to Develop (in Portuguese “Competências a Desenvolver”, designated by the acronym CD) are:

• CD1. Ability to formulate questions that can be answered by a KE/ML project
  
• CD2. Ability to analyse the feasibility, plan, and implement a KE/ML project
  
• CD3. Ability to propose, create, and interpret models suitable for real-world challenges
  
• CD4. Ability to critique current models and propose alternative models

• P1. Introduction to Knowledge Extraction (KE) and Machine Learning (ML)
  • 1.1 Initial Context
  • 1.2 Knowledge Discovery in Databases
  • 1.3 Knowledge Extraction versus Machine Learning
  • 1.4 Related Areas
  • 1.5 Tasks and Techniques
  • 1.6 Open Data
• P2. Categories of Machine Learning
  • 2.1 Supervised Learning
  • 2.2 Unsupervised Learning
  • 2.3 Reinforcement Learning
• P3. Predictive Activities
  • 3.1 Classification
  • 3.2 Regression
• P4. Descriptive Activities
  • 4.1 Clustering
  • 4.2 Summarization (and Visualization)
  • 4.3 Association
• P5. CRISP-DM Methodology
• P6. Main Techniques
  • 6.1 Decision Trees
  • 6.2 Association Rules
  • 6.3 Regression Linear with regularization
  • 6.4 Artificial neural networks
  • 6.5 Fuzzy sets and fuzzy logic
  • 6.6 Bayesian networks

The assessment process will comprise two components: a final written exam (EFE) and a practical project (PP). No continuous or periodic assessment is foreseen, only assessment by exam.

The Assessment Methods (in Portuguese, "Métodos de Avaliação", or the acronym MA) are as follows:

• MA1. Final Written Exam (EFE) (50%): Closed-book exam, which will take place in any of the regular exam periods. Taking the final written exam (EFE) component is mandatory in any assessment period; the grade obtained in a previous assessment period will not be carried over. Minimum passing grade of 7 (out of 20) points. It is mandatory.
• MA2. Practical Project (PP) (50%): It presupposes a practical project proposal to be submitted by the student(s), according to a specific proposed model, to be discussed and validated beforehand by the teacher. Presented and defended on the same day as the exam for the respective period, immediately following the final written exam. The practical project (PP) may be carried out individually or in a group of up to 4 members. The grade for the practical project (PP) component obtained in any assessment period will be valid for any subsequent assessment period, provided it occurs within the same academic year. The practical project (PP) component may be improved if the student takes an exam in a later assessment period. It is mandatory.

The Final Grade (CF) will correspond to the arithmetic mean of these two components. Final Grade Formula:

CF = 0.50 x MA1 + 0.5 x MA2

NAO