Logo

Final Project Submission

13 min read
Lesson slides

Machine Learning, Week 11

Final Project Submission

Put ten weeks of milestones together, hand in the project, and compete for the best one.

This is the last section of the course, and there is no new algorithm in it. Week 11 of the plan is the final submission of your team project: the code, a video, the proposal, the presentation and a logo, with a user interface as an extra. This short guidance session lists every deliverable, shows how the milestones of weeks 1 to 10 have already built most of it, and names the topics the plan leaves for you to explore after the course.

Objectives

By the end of the section you should be able to:

  • List the five required deliverables of the final submission and the one additional deliverable.
  • Say which weekly milestone produced each part of your submission.
  • Check your team's progress against the submission checklist and plan the work that is left.
  • Plan a project video that fits exactly in 5 minutes.
  • Name the three topics the plan suggests for future study: SVM, GBMs and multiple linear regression.

Where this sits in the course

Week 11 of the plan has no notebook and no dataset. It has two parts:

  • The project final submission. The plan lists six items: the project code on a GitHub repo, a project video of 5 minutes, the project proposal, the project presentation, the project logo and, additionally, a project user interface (mobile, web or desktop).
  • The course competition. The best projects are chosen as the winners of the course competitions.

Every earlier week ended with a project milestone. Those milestones were not separate homework: each one produced a piece of the final submission. Today you put the pieces together.

PartWhat we doTime
1The six deliverables15 min
2From weekly milestones to the submission10 min
3The competition and the topics for the future5 min
4Practice: your team's final check30 min

Part 1: The six deliverables

The checklist

#DeliverableWhat you hand inRequired
1Project codeA GitHub repository with your notebook and dataYes
2Project videoA video of 5 minutesYes
3Project proposalThe proposal your team wrote in week 2Yes
4Project presentationThe slides you presented in weeks 8 to 10Yes
5Project logoA logo for the projectYes
6Project user interfaceA mobile, web or desktop interfaceAdditional

1. Project code on a GitHub repo

The repository is where the reader checks that your results are real. It should hold:

  • The notebook, or notebooks, that run from top to bottom: loading the data (week 1), exploration and cleaning (week 3), features and the target variable (week 4), and the models (weeks 5 to 7).
  • The dataset, or a link to it when it is too large to upload.
  • A short README: the problem, the dataset, how to run the notebook, and the score of each model on the same test split.

Before you push, restart the runtime and run every cell from the top. Every number in your README, your slides and your video should be one the notebook prints.

2. Project video, 5 minutes

The video tells the story of the project to someone who did not attend your presentation. Five minutes is 300 seconds, so plan the time of every part before you record. The worked example below builds one plan.

3. Project proposal

You wrote it in week 2: the idea or the paper, the dataset with its rows, columns and target, and the first plots. Read it again and check that it still describes the project you built. If the dataset or the target changed since week 2, say so in the proposal.

4. Project presentation

You presented and discussed the project in weeks 8 to 10. Update the slides with what the discussion taught you and with your final scores. Keep the structure of the discussions: the idea, the data, the features and the target, the model and how you evaluated it, and what you would improve.

5. Project logo

The logo is the new item of week 11: no earlier milestone produced it. Use it on the first slide of the presentation, at the start of the video and at the top of the README, so the three look like one project.

6. Additionally: a project user interface

A mobile, web or desktop interface that lets a user try the model. It is not required, but it shows the model working on new input, which is exactly what a strong demo in the video needs.

Part 2: From weekly milestones to the submission

The milestones of weeks 1 to 10

WeekTopic of the sectionProject milestoneFeeds deliverable
1Advanced PandasTeam of 3 with a project idea, or 2 with a research paperAll of them
2Matplotlib and SeabornProject or paper proposal3 Proposal
3Linear regressionData exploration and data cleaning1 Code
4Logistic regressionFeatures extraction and target variable identification1 Code
5K-nearest neighborsMachine learning model implementation1 Code
6Decision trees and random forestMachine learning model implementation1 Code
7Naive BayesMachine learning model implementation1 Code
8k-means clusteringPresentations and discussions4 Presentation, 2 Video
9Content-based recommendationsPresentations and discussions4 Presentation, 2 Video
10Sentiment analysisPresentations and discussions4 Presentation, 2 Video
11Final submissionCode, video, proposal, presentation, logoAll of them

The interactive checklist

The checklist splits the five required deliverables into 13 items, and gives each item the week whose milestone produces it:

DeliverableItemsWeeks
1 CodeDataset loaded with pandas; data explored and cleaned; features and target chosen; models built (KNN, trees, Naive Bayes); code pushed to a GitHub repo1; 3; 4; 5 to 7; 11
2 VideoStory rehearsed in the presentations; a 5-minute video recorded8 to 10; 11
3 ProposalTeam with an idea or a paper; proposal written, with plots; proposal checked and included1; 2; 11
4 PresentationProject presented and discussed; final slides ready8 to 10; 11
5 LogoLogo designed11

The optional user interface is a 14th item, outside the 13.

Open the submission checklist widget full screen. Press Play to walk an on-track team from week 1 to week 11 and watch the progress ring fill, or tick your own team's items.

Worked example 1: how much is done by the end of week 10?

Question. A team met every milestone from week 1 to week 10. How many of the 13 required items does it already have when week 11 starts?

Step 1: the items finished in weeks 1 to 10. Go down the table above and keep every item whose last week is 10 or earlier:

WeekItems finished
1Dataset loaded with pandas; team with an idea or a paper
2Proposal written, with plots
3Data explored and cleaned
4Features and target chosen
7Models built (KNN, trees, Naive Bayes)
10Story rehearsed in the presentations; project presented and discussed

That is 2 + 1 + 1 + 1 + 1 + 2 = 8 items. Weeks 5 and 6 finish nothing on their own: the model item runs from week 5 to week 7, and it is done when the third model is.

Step 2: the fraction. 8 of 13 items is 8 / 13 = 0.6154, so the ring shows 62%.

Step 3: what is left for week 11. 13 - 8 = 5 items: push the code to a GitHub repo, record the 5-minute video, check and include the proposal, get the final slides ready, and design the logo.

The milestones built almost two thirds of the submission. Week 11 is for finishing, not for starting.

Worked example 2: a 5-minute video plan

Question. Split the 5 minutes of the video into five parts that follow the structure of your presentation.

Step 1: convert. 5 minutes = 5 x 60 = 300 seconds.

Step 2: give each part its time. The models are the core of the project, so they get the longest part:

PartSecondsFromTo
The problem and the data450:000:45
Cleaning, features and target600:451:45
Models and their scores901:453:15
A demo: the notebook, or the interface603:154:15
What you learned, what comes next454:155:00

Step 3: check the total. 45 + 60 + 90 + 60 + 45 = 300 seconds, exactly 5 minutes. Each part starts where the previous one ends.

Part 3: The competition and the topics for the future

The course competition

The plan chooses the best projects as the winners of the course competitions. A complete, honest submission is the base of a strong project:

  • All five required deliverables are handed in, and the interface is a bonus on top.
  • Every member of the team can explain every step, not only their own part.
  • The numbers in the slides, the README and the video are the ones the notebook prints.
  • The presentation names one thing that did not work and what you learned from it.

Additional points to explore in the future

The plan names three topics to explore after the course. Each one continues something you have already done:

TopicBuilds onscikit-learn class
Multiple linear regressionWeek 3: linear regression with more than one feature; the notebook of week 3 already predicts the price from five featuresLinearRegression
SVM (support vector machines)Weeks 4 to 7: another classifier used with the same fit, predict and scoreSVC
GBMs (gradient boosting machines)Week 6: another way to combine many decision trees into one modelGradientBoostingClassifier

If you try one of them on your project data, compare it with your other models on the same test split, as you did in weeks 5 to 7.

Common mistakes

  • Pushing a notebook whose cells were run out of order, so its printed outputs do not match what it computes.
  • Scores in the video or the slides that differ from the scores in the notebook.
  • A video far over or under 5 minutes because the parts were never timed.
  • A proposal that still describes the dataset or the target of week 2 after the project changed.
  • Leaving the logo, the one item with no earlier milestone, to the last hour.

Key takeaways

  1. The final submission has five required deliverables: code on a GitHub repo, a 5-minute video, the proposal, the presentation and the logo. A user interface is additional.
  2. The milestones of weeks 1 to 10 produced most of it: an on-track team starts week 11 with 8 of the 13 checklist items done.
  3. Week 11 finishes five items: push the code, record the video, check the proposal, finish the slides and design the logo.
  4. Time the video in seconds: 300 seconds in total.
  5. After the course, explore SVM, GBMs and multiple linear regression on your own data.

Practice

About 30 minutes. Try each task before you read its answer at the end of the page.

Practice 1: name the milestone week (about 5 minutes)

For each piece of work, name the week whose milestone produced it:

  1. Naming the target variable of your project.
  2. Choosing K for KNN with the elbow loop.
  3. Comparing the training score and the test score of a decision tree.
  4. Choosing between MultinomialNB and GaussianNB for your features.
  5. Stating your idea in two or three sentences and naming the dataset.

Practice 2: time a video (about 5 minutes)

A team has planned four parts of its video: 30 seconds, 75 seconds, 105 seconds and 50 seconds. The fifth and last part is the demo.

  1. How many seconds are left for the demo if the video must be exactly 5 minutes?
  2. At what time, in minutes and seconds, does the demo start?

Practice 3: your team's final check (about 20 minutes)

  1. Open the submission checklist widget and tick every item your team has really finished.
  2. Write down the items that are still missing.
  3. Give each missing item one owner in the team and a date before the submission deadline.
  4. Write your own 5-part video plan and check that the parts add up to 300 seconds.

Answers

Answer 1

Piece of workWeekMilestone
Naming the target variable4Features extraction and target variable identification
Choosing K for KNN with the elbow loop5Model implementation, with KNN
Training score against test score of a tree6Model implementation, with decision trees and random forest
MultinomialNB or GaussianNB7Model implementation, with Naive Bayes
The idea and the dataset in a few sentences2Project or paper proposal

Answer 2

  1. The four parts use 30 + 75 + 105 + 50 = 260 seconds, so the demo gets 300 - 260 = 40 seconds.
  2. The demo starts after 260 seconds, that is 4 minutes and 20 seconds, at 4:20, and ends at 5:00.

Answer 3

There is no single answer: it depends on your team. As a reference, a team that met every milestone starts week 11 with 8 of the 13 required items, and the 5 left are the code pushed to the repo, the video, the checked proposal, the final slides and the logo. Any other missing item comes from an earlier milestone, so give it an owner first. Your video plan is correct when its parts add up to exactly 300 seconds.