Wiki | Lisbon Data Science Academy logo Wiki | Lisbon Data Science Academy

Structure

The capstone project is a chance to put everything you’ve learned in practice. The Capstone must be completed individually and it is composed of two important work stages, each ending with the delivery of a report and an app. At the end of each stage there will be a comment round in the reports and a chance for the students to improve them with the provided feedback. In between stages there is a test round for your app, where real data will be sent to it. This data should be used by you for evaluation and eventually retraining of your model.

Stage 1

In the first part of the project, you will receive a client briefing containing the instructions to the problem you have to solve. This will contain a problem description, the available data to solve it and a list of requests:

All of which are to be delivered at the end of this stage.

App trial round

Close to the deadline for report 1 and the application, we will provide an interval where you can try out your app. This means we’ll use the procedure to be used on the actual test round to send dummy data to your app, so you can:

  1. check if it’s working
  2. if not, debug why
  3. check it’s performance (how many requests you were able to answer, for example)

This means that ideally you should have your app ready by then. This is not mandatory, as it is not an evaluation moment, but it is definitely recommended.

App Test round

Once you deliver your app and the first report, it’s game on! We’ll start sending requests into your app, in two rounds:

Don’t worry if you don’t understand right now what this will mean, you’ll get there. All you need to know is that at this point you must deliver a working app, i. e., it should return some prediction for our data.

Stage 2

In the second part of the project, you will analyse the real data you received and produce a second report going through it, retraining your model and redeploying it.

At the end of this you will produce a second report and will have a smaller app test round. Redeploying your model is not mandatory, but recommended to show improvements in your results.

Comment rounds

In each stage you produce a version of the report. This is an opportunity for you to receive feedback and improve your reports.

Time commitment

You should be prepared to spend a recommended 10 to 20 hours per week on the capstone, especially in its first 3 weeks.

Capstone Evaluation Rules

In order to pass the capstone all the following must be delivered and pass a set of standards:

API

All students should have delivered a working API by report 1 to pass. This will be verified by the following:

Report

In order to get a passing chance, both reports need to be sent with all sections completed. In addition, each section has to pass a minimum threshold of quality, with respect to the requirements presented. You can refer to the sections and descriptions of batch 4 to understand better how these may look.

Automatic fails

The capstone will automatically be considered a fail if:

Calendar

Description Date
Kick off 2023-04-03
Capstone Clarification email 2023-04-09
Trial round of requests 2023-04-23
Deadline Provisory report 1 and app launched 2023-04-30 , 23h59 UTC
First round of requests 2023-05-01 to 2023-05-07
Comments to report 1 made by instructors 2023-05-07 23h59 UTC
Deadline report 2 + redeploy + address comments to report 1 2023-05-28, 23h59 UTC
Second round of requests 2023-05-29 to 2023-06-02
Comments to report 2 made by instructors 2023-06-04 , 23h59 UTC
Deadline address comments to report 2 2023-06-11, 23h59 UTC
Graduates announced 2023-06-19