Machine Learning approach to Multidimensional Opinion Polarization - Informatics Engineering Specialization in Computing

Aug 4, 2026

Upload by Finn

Machine learning pipeline for analyzing multidimensional opinion polarization in survey data, framed within an informatics engineering specialization in computing. The work defines context, scope, project objectives, and methods, then preprocesses data by selecting useful and shared variables, ensuring consistency and handling missing values. It learns a low-dimensional embedding of respondents via dimensionality reduction (including ISOMAP), forms topic clusters reflecting opinion, and evaluates results with statistical tests comparing political and socio-demographic attributes. Finally, it presents temporal planning, resources, risks, budget, and sustainability dimensions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages88

Reading Time3 h 42 min

AccessFree to download

Context and scope

Introduction

Contextualization

Scope

Methodology and tools

Methods

Data preprocessing

Low dimensional embedding of the respondents

Cluster of topics forming an opinion

Results

Low dimensional representation of individuals in the space of topics/questions

Topic clusters representing an ideology

Conclusions

Temporal planning

Description of tasks

Necessary resources

Time estimate

Gantt

Risk management

Budget

Personnel costs

Generic costs

Contingencies

Total cost

Sustainability

Environmental dimension

Economic dimension

Social dimension

References

Annex A: Chosen variables

Annex B: Low dimensional representation of different sets of years

Annex C: Effect of ISOMAP hyperparameters

Annex D: Handpicked clusters

Machine Learning approach to Multidimensional Opinion Polarization Informatics engineering Specialization in Computing Author: David Cárcamo Delgado Director: Romualdo Pastor Satorras Co-director: Michele Starnini GEP tutor: Joan Subirats Soler January 23, 2023 Contents 1 Context and scope 10 1.1 Introduction ................................. 10 1.1.1 Implied Actors ........................... 10 1.2 Contextualization .............................. 11 1.2.1 Previous studies ........................... 11 1.2.2 Justification ............................. 12 1.3 Scope .................................... 12 1.3.1 Project objectives and sub-objectives ............... 12 1.4 Methodology and tools ........................... 13 1.4.1 Methodology ............................ 13 1.4.2 Tools ................................. 13 2 Methods 15 2.1 Data preprocessing ............................. 15 2.1.1 Dataset characteristics ....................... 15 2.1.2 Selection of useful variables .................... 15 2.1.3 Selection of shared variables .................... 17 2.1.4 Variable consistency ........................ 20 2.1.5 Missing Data ............................ 21 2.2 Low dimensional embedding of the respondents ............. 23 2.2.1 Dimensionality reduction ...................... 23 2.2.2 Dimensionality reduction techniques explored .......... 24 2.2.3 Creation of the embedding ..................... 25 2.2.4 Comparison of political and socio-demographic attributes .... 25 2.2.5 Statistical tests ........................... 26 2.3 Cluster of topics forming an opinion .................... 28 2.3.1 Creation of clusters ......................... 28 2.3.2 Comparison of clusters ....................... 30 3 Results 31 3.1 A low dimensional representation of individuals in the space of topics/questions ................................. 31 3.1.1 Effect of political leaning on ideology ............... 32 3.1.2 Comparing other socio-demographic attributes .......... 34 3.1.3 Low dimensional representation of different sets of years .... 38 3.1.4 Statistical tests ........................... 38 3.1.5 Effect of ISOMAP hyperparameters ................ 40 3.2 Topic clusters representing an ideology .................. 41 3.2.1 Study of obtained clusters ..................... 41 4 Conclusions 45 5 Temporal planning 46 5.1 Description of tasks ............................. 46 5.1.1 Project management (GP) ..................... 46 5.1.2 Data preparation (PD) ....................... 47 5.1.3 Dimensionality reduction (RD) .................. 47 5.1.4 Clustering (C) ............................ 47 5.1.5 Results study (ER) ......................... 48 5.2 Necessary resources ............................. 48 5.2.1 Human resources .......................... 48 5.2.2 Material resources ......................... 48 5.2.3 Software resources ......................... 49 5.3 Time estimate ................................ 49 5.4 Gantt .................................... 51 5.5 Risk management .............................. 52 6 Budget 53 6.1 Personnel costs ............................... 53 6.2 Generic costs ................................ 55 6.2.1 Amortizations ............................ 55 6.2.2 Energy consumption ........................ 55 6.2.3 Internet consumption ........................ 56 6.2.4 Total generic costs ......................... 56 6.3 Contingencies ................................ 56 6.4 Unforseen .................................. 57 6.5 Total cost .................................. 58 7 Sustainability 59 7.1 Environmental dimension .......................... 59 7.2 Economic dimension ............................ 59 7.3 Social dimension .............................. 59 8 References 61 9 Annex A: Chosen variables 64 10 Annex B: Low dimensional representation of different sets

FAQs

What dataset handling steps are included before modeling?

The approach preprocesses data by describing dataset characteristics, selecting useful variables and shared variables, checking variable consistency, and addressing missing data.

How is the low-dimensional representation of respondents obtained?

Respondents are embedded using dimensionality reduction techniques, including creation of the embedding and comparisons against political and socio-demographic attributes, supported by statistical tests.

How are opinion topics and ideologies identified in the results?

The method clusters topics to form clusters representing opinions, then compares obtained clusters and studies them to interpret ideologies, using measures such as similarity and mutual information.

Aug 4, 2026

Upload by Finn