• 

  Generator component

• 

  Overall Architecture

• 

  Algo component

• 

  Comparison of average utilization_rate for different algorithms.

• 

  Comparison of matching_loss for different algorithms.

• 

  Comparison of utilization_rate histogram for different algorithms.

Our project

Abstract

Assigned team number

Please refer to our GitHub repo for the latest update.

Chosen Problem statement:

We choose Integrity Data Science(Optimize advertisement moderation)

What inspired us:

We all share a strong interest in the hackathon and the selected topic, and we are eager to expand our expertise through this project.

Development tools used to build the project

Libraries used in the project:

"numpy", "pandas", "scikit-learn", "matplotlib",

Accomplishments that we're proud of

Functional Features:

An array of Algorithms that estimate values and allocate advertisement to moderators in run time.

Simulate advertisement arriving at different patterns (peak hours, uniform distribution etc).

Analysis of the properties distribution pattern in the data input to generate "virtual" advertisement/moderator objects

And of course, we also support sampling from the real advertisement/moderator inputs

A Simulator that combines all the components above to produce analysis of the behaviors of each algorithm under particular input.

A Visualization component and help with the analysis of the result of simulation.

What we learned

1. We discovered like-minded friends, and the project became an enjoyable and rewarding challenge.
2. Through our experience in optimizing advertisement moderation, we've gained valuable insights that will prevent future confusion when encountering similar issues.
3. We extensively researched various materials and methods, significantly enhancing our professional skills along the way.

What challenges we faced

1. Given the diverse backgrounds of our group members, our group's inception occurred amidst time constraints. Furthermore, we grappled with the need to effectively allocate our time between academic obligations and project endeavors.
2. Initially, the project's architectural framework remained shrouded in ambiguity, leaving us uncertain about how to initiate our work. It was only through several productive discussions that each of us gained a clearer understanding of our individual roles and responsibilities.
3. Of utmost importance, a substantial portion of the project involves technical complexities that surpass our individual expertise. Consequently, we frequently engage in collaborative discussions to devise effective algorithms and refine data processing methods.

What's next for AllocAvengers

AllocAvengers' future hinges on personalization through AI, targeting individual users with tailored content. AI-driven creativity will simplify ad creation, and cross-channel integration will expand brand reach. User-centricity, data privacy, and advanced attribution modeling will continue to shape ethical advertising practices, fostering innovation and personalization.

Meet the team

A team from National University of Singapore, we come from diverse subject backgrounds.

Junwu: A mathematics and computer science student interested in machine learning and algorithms.

Ziwen: A hacker keen on exploring new things.

Yiwen: A Second year computer science student second majored in mathematics.

Zebang: A second year student majored in Data Science & Analytics.

Shuyao: A year 2 CS major.

Built With

• python