University of Oulu Business Analytics CSE Master's Thesis Guide: 521029S, 30 ECTS, Examiners, 521009S Maturity Test and Laturi

Quick answer: what is the University of Oulu Business Analytics CSE thesis route?

The current Business Analytics - Computer Science and Engineering path at the University of Oulu leads to a Master of Science (Technology) degree. The programme is a two-year, 120 ECTS international Master’s route, and the current 2026-2027 Peppi programme object is 52492 / IMP2026BACSE. The exact thesis is 521029S Master’s Thesis, Business Analytics (Computer Science and Engineering), 30 ECTS, Peppi object 19773, classified as Advanced Studies, taught in English and graded 1-5/FAIL. The same current programme structure requires 521009S Computer Science and Engineering, The Maturity Test for Master’s Degree, 0 ECTS, assessed pass/fail.

The thesis process is different from the Business path of the same Business Analytics programme. For the CSE route, current University instructions establish a Laturi-based research-plan and supervision process, a supervisor plus second examiner evaluation model, published BA-CSE score ranges and Degree Programme Committee approval. Current programme 52492 does not establish a separate credited thesis-seminar course, so this guide does not invent one. The current official realization endpoint also does not expose a 521029S realization, so no unverified implementation code or date is stated.

1. Confirm that you are in the CSE degree path

Business Analytics has three degree-specific paths. The CSE path awards Master of Science (Technology) and is oriented toward Data Engineer competence. The Business path awards MSc in Economics and Business Administration, while the Software Engineering and Information Systems path is another separate guide object. Shared Business Analytics courses do not make the thesis routes interchangeable. Before using thesis instructions, confirm that your Peppi programme is 52492 Business Analytics, Computer Science and Engineering (MSc., tech), International Programme 2026-2027.

2. The exact current thesis is 521029S and it is 30 ECTS

Current programme 52492 places 521029S Master’s Thesis, Business Analytics (Computer Science and Engineering) inside the Master’s Thesis and Maturity Test module. The thesis is exactly 30 ECTS. The live current course object is 19773, the level is Advanced Studies, the primary teaching language is English and the assessment scale is 1-5/FAIL. These programme and course records should control current planning. Do not substitute 721020S from the Business path or 522987S from Biomedical Engineering.

3. Do not invent a current thesis implementation code

The official current Peppi course object confirms 521029S, but the current official realization endpoint does not expose a separate realization for the thesis. This is a useful evidence boundary. A guide should not create an implementation code, teaching period or registration date simply because many other courses have those fields. Use 521029S as the verified thesis object and check the live Peppi/Laturi environment for any administrative implementation information that becomes available later.

4. The current programme does not show a separate credited thesis seminar

Programme 52492 contains a Master’s Thesis and Maturity Test module of 30 ECTS with 521029S thesis and 521009S maturity test. Current structure does not establish an additional credited BA-CSE thesis seminar. University guidance can refer to programme-dependent seminar activity, but that is not enough to create a separate ECTS object. If your supervisor organises presentations or group sessions, follow them, but do not report extra seminar credits unless a current official programme object establishes them.

5. Start the thesis process in Laturi

Current BA MSc Technology thesis instructions use Laturi as the formal workflow. Laturi is used for starting the thesis, supervision, monitoring, evaluation, plagiarism checking and publication. Treat the system as part of the academic process rather than only a final upload page. The thesis should be set up there with the correct programme, supervisor information and research plan before the work advances to final evaluation.

6. Prepare the research plan before building the full study

For the BA MSc Technology route, the research plan is currently described as a maximum of one A4 page. It should identify the purpose of the work, the main methods, materials, devices or software, and the expected results. A one-page limit requires precision. State the problem, data or technical material, intended analytical or engineering method, evaluation logic and expected contribution. Do not fill the page with broad background that belongs later in the thesis.

7. Make the research question technical and researchable

The CSE Business Analytics route combines technical data competence with business and organisational relevance. A useful thesis question should therefore be narrow enough for technical evaluation and meaningful enough to justify the work. Questions such as whether one predictive approach generalises better than another, whether a data-processing architecture improves a defined performance constraint, or whether a model supports a decision task under specified conditions are usually easier to evaluate than a broad question such as “How can AI improve business?”

8. Match the method to the claim you want to make

A thesis can include machine learning, statistical modelling, data pipelines, dashboards, NLP, big-data processing or another technical solution. The method should be chosen because it can answer the research question, not because it is fashionable. If the claim is predictive, the evaluation must address unseen data. If the claim concerns system performance, the experiment must measure the relevant technical dimensions. If the claim concerns users or organisational outcomes, technical accuracy alone may not be sufficient evidence.

9. Current CSE instructions define who may supervise

The current CSE thesis instructions state that the supervisor can be a CSE, EE or DCE professor, or a postdoctoral researcher currently working at CSE, subject to the programme’s qualification conditions. A postdoc acting as supervisor must have previous experience as a second examiner. Do not assume that any company mentor or project manager can act as the formal University supervisor. External experts can be important, but formal roles must meet the current CSE requirements.

10. A second examiner is part of the current CSE route

Current instructions establish a second examiner in addition to the supervisor. The second examiner may be a CSE professor, a professor from another field or a postdoc currently working at CSE, again subject to current eligibility conditions. The supervisor and second examiner evaluate the thesis in Laturi. This is a path-specific rule and should not be replaced by the Business path’s OBS process.

11. Check the senior-qualification condition early

The current CSE process includes a qualification condition concerning the supervisor and second examiner. In practice, either the supervisor or the second examiner must satisfy the stated senior qualification requirement in the current CSE instructions. Because formal eligibility can affect Laturi setup and evaluation, confirm the examiner combination before the thesis is close to submission. Do not wait until the final week to discover that the intended examiner pair does not satisfy the programme rule.

12. Use the kick-off discussion to set the thesis contract

Current CSE guidance calls for an initial discussion covering the topic, schedule, supervision implementation and assessment criteria. Use this meeting to define practical expectations. Agree how often drafts will be reviewed, what form of evidence is expected, how company data will be handled, who can approve changes in scope and what counts as a sufficiently complete manuscript. A clear kick-off reduces later disagreement about what the thesis was supposed to deliver.

13. A technical supervisor is required for work outside the University

When the thesis is carried out outside the University, current CSE instructions require a technical supervisor. This person can support the practical or engineering side of the work in the host organisation. Keep the roles clear: the technical supervisor can provide domain and implementation guidance, while formal University supervision and examination must still follow the CSE academic process. This distinction is especially important in company-sponsored theses where organisational objectives may not fully match academic requirements.

14. Company-sponsored work still needs an academic research contribution

The current CSE route supports company and sponsored thesis work, including current CSE-ITEE opportunities. However, a useful product, model or software feature is not automatically a Master’s thesis. The manuscript should explain the scientific or technical problem, relevant prior work, method, evaluation criteria, results and limitations. A company can define the practical problem, but the thesis must convert it into a defensible academic investigation.

15. Use programme courses to understand the expected methods environment

Current programme 52492 includes Data Analytics and Business Intelligence, Towards Data Mining and Statistical Methods for Business Analytics in its main study environment. It also offers CSE specialisation options such as Machine Learning, NLP and Text Mining, Big Data Processing, Deep Learning, Multi-Modal Data Fusion and AI Ethics, Privacy and Legislation. These courses do not dictate one thesis method, but they show the analytical and technical environment in which the thesis is expected to operate.

16. Data analytics requires a complete chain from collection to communication

The current Data Analytics and Business Intelligence course covers gathering relevant data, preparing and modelling data, analysing it, visualising findings and communicating results. Use the same chain in the thesis. A technically correct model is not enough if data provenance is unclear or findings cannot be interpreted. Document where the data came from, how it was transformed, what model or analysis was applied and how the final result answers the research question.

17. Data preprocessing is part of the method, not a hidden preparation step

Towards Data Mining explicitly covers data collection design, combining data sources, normalisation, transformations, missing or incorrect values and generalisability. These steps can strongly affect results. Report important preprocessing decisions in the Methods chapter and preserve them in reproducible code where possible. If thousands of rows disappear during cleaning, explain why. If a variable is transformed or imputed, record the rule and its justification.

18. Separate training, tuning and final evaluation

Current programme methods explicitly teach train-test-validation and other validation strategies. For predictive work, do not use the same data repeatedly for feature selection, hyperparameter tuning and final performance reporting. Development choices should be made using training and validation information, while the final test data should remain outside those decisions if it is intended to estimate generalisation. Otherwise the reported result can become optimistically biased.

19. Cross-validation should respect the real data-generating unit

Standard random k-fold cross-validation is not automatically correct. If the dataset contains many records from the same customer, patient, device, company or user, splitting rows randomly may place closely related observations in both training and validation sets. Decide what the model is expected to generalise to and group the evaluation accordingly. A thesis should explain why its resampling design represents the intended deployment or inference problem.

20. Time-dependent data require chronological thinking

Business and operational datasets often include time. If the thesis predicts future sales, failures, churn or demand, random splitting can allow future patterns to influence training. Where chronology matters, build the evaluation so that development data precede evaluation data. Explain any rolling-window, temporal holdout or backtesting design. The goal is not to use the most complicated validation scheme, but to avoid giving the model information that would not exist at deployment time.

21. Statistical analysis should discuss validity and generalisability

721026S Statistical Methods for Business Analytics explicitly covers usability, validity, reliability and generalisability of business data and reports. Carry these questions into the thesis. Are variables valid measures of the intended concepts? Is the sample representative of the claim? Could missing observations create selection bias? Are repeated observations independent? Does uncertainty remain acceptable? A p-value or coefficient is not meaningful unless the data and design support its interpretation.

22. Machine learning needs more than one headline accuracy number

The current Machine Learning course includes regression, classification, optimisation, feature engineering, validation, kernels, neural networks and tree-based methods. Choose evaluation metrics according to the problem. Accuracy can be misleading with imbalanced classes, and average error can hide important subgroup failures. Report denominators, class balance and relevant baseline models. Where appropriate, include precision/recall, ROC or PR measures, calibration, MAE/RMSE or other task-specific metrics.

23. Error costs should reflect the business or technical decision

Two models with similar overall performance can create very different consequences. A false positive may waste resources while a false negative may miss a critical event. Define the practical meaning of prediction errors before choosing thresholds or a preferred model. If the thesis recommends deployment, connect technical performance to the decision context. Do not imply that the statistically best model is automatically the most useful operational model.

24. NLP and text-mining theses need corpus provenance

The current NLP and Text Mining environment covers retrieval, text categorisation, corpus inference and language-processing tools. If your thesis uses text, document how the corpus was collected, which languages it contains, what was excluded, how labels were created and how duplicates or leakage were controlled. If large language models or embeddings are used, record the model/version or service context where possible and explain how outputs were evaluated against the research question.

25. Deep-learning complexity is not evidence of quality

The current Deep Learning course covers CNNs, RNNs, attention, transformers, GANs, diffusion models and practical model training. A thesis should choose a deep model only when the problem and data justify it. Compare against meaningful simpler baselines where possible. Report training configuration, validation logic and limitations. A complex architecture with no controlled comparison is harder to interpret than a simpler model with a clear research design.

26. Big-data scale does not remove research-design limitations

Big Data Processing covers data-intensive systems, storage, batch and stream processing, privacy, security and analysis. Large row counts can improve some technical tasks, but they do not automatically create a representative sample or independent observations. A billion events from one platform can still represent a narrow population. Keep population, data-generation process, access restrictions and evaluation unit explicit when making general claims.

27. Multi-modal data fusion needs alignment and source-specific provenance

Current Multi-Modal Data Fusion studies include combining different data sources, alignment, Bayesian inference and machine-learning fusion. If the thesis combines text, images, sensors, transactions or other modalities, explain how observations are aligned and what happens when one source is missing. Preserve provenance for each source. Fusion can create strong models, but it can also hide errors when one source is systematically noisier or collected under different conditions.

28. AI ethics and privacy should influence design choices

The current AI Ethics, Privacy and Legislation course asks students to examine ethical and legislative conditions in AI development and deployment. For a thesis, consider whether the model affects people, whether sensitive attributes are involved, whether automated decisions create unequal risks, whether training data were lawfully obtained and whether outputs can be explained sufficiently for the context. Ethics should shape design and evaluation rather than appearing only as a short paragraph at the end.

29. Personal-data planning must happen before processing

University data-protection guidance requires the lifecycle of personal-data processing to be documented in the research plan before processing begins. The plan should address data protection and information security, include a risk assessment and consider a DPIA where applicable. Research subjects must receive an appropriate privacy notice where required. Collect only data necessary for the academic purpose and anonymise or pseudonymise where possible.

30. Pseudonymised data can still be personal data

Replacing names with IDs does not automatically make data anonymous. If a key exists or a person can reasonably be re-identified using other fields, the data remain personal data. Technical datasets may contain indirect identifiers such as timestamps, device IDs, locations, job roles or rare behavioural patterns. Keep re-identification keys separate, limit access and avoid moving personal data into unapproved tools or cloud services.

31. Ethics review depends on the actual research design

Not every BA-CSE thesis requires an ethics-committee statement. However, projects involving human participants, sensitive personal information, interventions, deception or other applicable criteria may require preliminary ethical assessment. The University Human Sciences Ethics Committee handles relevant non-medical human-sciences requests, and Master’s students need the required supervisor involvement. Classify the study early so ethics review, if needed, occurs before the relevant research activity.

32. Build a data-management plan that supports reproducibility

University responsible-research guidance expects planning for collection, storage, sharing, preservation and reproducibility. A practical thesis data-management plan can include data inventories, source and licence information, storage locations, access rights, variable dictionaries, preprocessing scripts, model configuration, output folders and retention/deletion decisions. Company or licensed data may remain closed, but the analytical provenance should still be documented so the result can be audited responsibly.

33. Research integrity applies to code and AI-assisted work

Current University integrity rules apply beyond copied prose. A technical thesis can also have attribution problems in code, notebooks, datasets, model outputs and AI-assisted material. Keep track of external repositories, packages, generated code, prompts or generated text when they materially contribute. Follow the current University instructions for acknowledging AI use. Similarity checking cannot prove that code, data or analytical choices are original, valid or properly attributed.

34. The thesis is evaluated by supervisor and second examiner

In the current CSE process, the supervisor and second examiner discuss and agree on the evaluation. The supervisor writes the evaluation in Laturi and the second examiner joins the evaluation. Current CSE instructions publish the BA-CSE score mapping as 10-15 = grade 1, 16-21 = 2, 22-28 = 3, 29-35 = 4, and 36-42 = 5. Use the current assessment criteria throughout drafting rather than reading them only after the thesis is finished.

35. Degree Programme Committee approval is part of the final assessment

After the examiners complete their evaluation, the thesis grade is accepted through the Degree Programme Committee process. Current instructions also require the relevant plagiarism/similarity result and evaluation proposal to be handled before committee approval. This means the final schedule should include more than the student’s upload date. Allow time for supervisor review, second examiner participation, evaluation and committee handling.

36. Know the rectification boundary

Current CSE and University assessment rules provide a rectification route after the student is informed of the assessment. The current CSE instructions refer to a 14-day period in the applicable process. Read the current decision and assessment instructions immediately if you believe there is an error. An approved and graded thesis cannot simply be treated as a normal assignment that can be resubmitted for a higher mark.

37. The maturity test is 521009S and carries 0 ECTS

Programme 52492 requires 521009S Computer Science and Engineering, The Maturity Test for Master’s Degree, 0 ECTS, assessed PASS/FAIL. Current 521009S course information says the maturity test is evaluated and approved by the thesis supervisor and describes a supervisor-provided topic in a controlled written event. At the same time, the University’s current ITEE faculty guidance says the maturity test is primarily completed as an electronic exam in Examinarium, and the live 521009S-3005 implementation is classified as Scheduled Online. Treat the delivery mode as an implementation detail to confirm with the supervisor and the current Peppi entry before sitting the test; the 521009S code, 0 ECTS and PASS/FAIL requirement remain stable.

38. Current maturity guidance gives an indicative length of 450-600 words

The current 521009S course description still gives an indicative length of about three handwritten pages or 450-600 words. Because the current ITEE faculty guidance primarily points students to an electronic Examinarium exam, treat this length and handwritten wording as course-description guidance rather than proof of the physical test format for every current sitting. Confirm the active Peppi implementation and delivery details with the supervisor. The purpose remains to demonstrate familiarity with the thesis field and the applicable language requirement; do not assume that an English-taught programme makes the maturity-test language automatically English for every student.

39. Laturi submission must contain a public-safe thesis

Laturi handles final evaluation and publication. The final thesis must not contain secret trade or professional information. If a company project involves confidential background data, agree the handling before writing sensitive details into the manuscript. The public thesis should contain enough information to support its academic conclusions without disclosing material that cannot legally or contractually be published. Do not wait until final submission to discover that core figures must be removed.

40. Plan graduation after thesis and maturity-test completion

When the thesis, maturity test and other required studies are approved, degree application proceeds through Peppi. Current graduation instructions provide BA MSc Technology scheduling information and note that Degree Programme Committee meetings can also be arranged outside announced schedules when necessary. Because public timetable details can change, use the live graduation page when choosing a target date instead of relying on an old screenshot or a copied calendar entry.

41. A practical 12-step BA-CSE thesis workflow

1. Confirm programme 52492 / IMP2026BACSE. 2. Confirm 521029S / 30 ECTS as the thesis object. 3. Identify an eligible University supervisor and, where applicable, a technical supervisor. 4. Complete the kick-off discussion. 5. Prepare the maximum-one-A4 research plan and start the process in Laturi. 6. Resolve data access, confidentiality, personal-data and ethics requirements before processing. 7. Build a reproducible data and analysis pipeline. 8. Separate development from final evaluation. 9. Write the full thesis with method, results, limitations and defensible conclusions. 10. Complete supervisor and second-examiner evaluation and committee handling. 11. Complete 521009S / 0 ECTS maturity test. 12. Finalise Laturi publication and apply for the degree through Peppi.

42. Final pre-submission checklist

Verify that 521029S remains the current 30 ECTS BA-CSE thesis and 521009S remains the current 0 ECTS maturity test. Confirm supervisor and second-examiner eligibility, technical-supervisor arrangements if the work is outside the University, Laturi research-plan approval, current assessment criteria, data and ethics permissions, reproducible preprocessing and model evaluation, public-safe company information, maturity-test language and current graduation timetable. Finally, recheck the curriculum boundary. The University is transitioning to the 2027-2030 curriculum, so current 2026-2027 thesis details should be freshly verified before they are treated as final rules for autumn-2027 study arrangements.