Project Presentation: Project Realice

Team Johnson-Sirleaf

Our visualization project is about finding solutions to Police Brutality, a contemporal challenge that many young people in Nigeria have been exposed to. The amazing team of young Data Scientist (TEAM JOHNSON-SIRLEAF), of AI Saturdays Lagos Cohort-7 learners publish their findings and solutions with one goal in mind — setting a foundation that others can build on to bring about a more inclusive solution towards Policing and its Reformation in [Nigeria].

Table of Contents

1. Recanting the Learning Experience

2. Aims & Objectives

3. Data Collection

4. Data Preprocessing & EDA

5. Selection: Selecting Seeds from Chaffs

6. The End Game: Visualize, Visualize, Visualize

7. Deployment

8. Future Development

9. Conclusion

Recanting the Learning Experience

It is essential to mention that on the road to building this solution, we had covered the following in 16 weeks:

• 16 weeks of Theoretical Classes
• 10 Week of Practical Projects
• and of course, the Capstone project

The organizers documented the entire learning experience here.

Aims and Objectives

Photo by Engin Akyurt on Unsplash

For every task, a set objective is expected, as this help to drive the actions of the teams; ours wasn’t any different. With Project Realice, we wanted to tell a complete story of Policing in Nigeria. Here is a quick peek at our objectives:

1. A data pipeline (databases) of Police harassment incidence.
2. Mapping of Police harassment per state
3. A word-cloud description of the Police in Nigeria
4. A periodically updated dashboard scoring system that gauges public confidence in Police activities.

A little bad news**: objective no [4] was not achieved for many apparent reasons, including time, resources and expertise. Hopefully, our team fixes this with more collaboration, especially from you reading this.

Data Collection

Collecting diverse opinions can be hard, and we found no better way of collecting anonymous public data than via google forms. This helps us present it in the most straightforward manner to the targeted audience. Initially, we set forth to get over 100 000 respondents by leveraging on all our social media platforms, but as time passed, the team realized we had to settle for 0.2% of the initial target.

Data Preprocessing /EDA

Next — so after missing out on the big target, we proceeded to export the responses we got so far and start with some basics wrangling:

• Firstly, removing sensitive responses like email, sex etc. The age range was not removed.
• Then came the handling of missing data.
• After which, we reformatted error input by responders for each of the 30+ features/columns.
• then we finally checked for impossible scenes and headed to the next phase

This was very quick since the dataset was not so large.

Feature Selection: Selecting the Seeds from Chaffs

At this phase of the project, we decided to select the top 8 features that can help us tell the true story of Police activities. From the correlation coefficient, we found out the following features we the gem:

• age range
• frequency of encounter positive encounter
• frequency of negative police encounter
• general frequency count(always, seldom, rarely)
• police unit encountered count
• year(s) of harassment
• one sentence advice on police activities
• finally, one word to describe the police

The End Game: Visualize, Visualize, Visualize

When you realize that the end game is near you start rounding off. The team did an excellent job of cross visualizing the following features randomly against each other to see what it produces.

Well, it happened that matplotlib wasn’t going to be enough to help create the visualizations we needed, so we moved to use Plotly, another python visualization library, to help us achieve the aim of different charts.

Deployment: A Demo

After completing the project and pushing it to our repo, we decided it was best to deploy it via streamlit. A demo of our findings can be found here:

https://share.streamlit.io/aisaturdayslagos/cohort7-team-johnson-sirleaf/main/Project/app.py

Future Development

So far, we have presented our project to you. However, we intend to do more for this project as a team (writer’s view).

• Make the data publicly available. For now, reach out to our team on (mail).
• Collect as many as 1 million responses.
• Work on achieving Objective [4].
• Hosting the improved versions of this project on our server.
• Getting authorities involved to collaborate with us

You, too, can help us by filling out the forms or leaving a suggestion in the comment.

Conclusion

Here are some lessons we have learnt as a team:

• User all possible means and channels to collect data (Meta, Twitter, TikTok, Instagram).
• Always choose workability over aesthetics.
• Bugs are your friends until you don’t want them anymore.
• A project can be finished, but maintenance and upgrade keep the solution relevant.

Shoutout to my Teammates — Abraham, Margaret, Peace, Paul, Oladev

Team Mentors — Tejumade Afonja, Yorubayesian, kaaee.