Visitor Prediction + Micro-Targeted Ad Campaign

Click Here to check the matrix, zip code excels and complete case study

To help the National Center for Civil and Human Rights (NCCHR) increase local visitor attendance, I designed two complementary AI-powered projects:

1. Visitor Prediction Model — to identify individuals most likely to visit based on digital behavior.

2. Micro-Targeted Ad Campaign — to deliver hyper-personalized social media content by ZIP code using AI-generated copy and visuals.

   
   

Together, these initiatives offer a scalable framework to turn awareness into real attendance — with limited budget and high community relevance.

What I Did?

​

I built a machine learning model that predicts whether a person is likely to visit NCCHR based on their online actions — like visiting the website, searching for discounts, or reading online reviews.

​

Why I Did It?

​

NCCHR has strong brand awareness, but low conversion. By understanding what behaviors correlate with real visits, we can focus outreach on those with high intent — increasing marketing efficiency and impact.

​

How I Did It?

​

Methodology:

​

• Cleaned and anonymized student behavior data

• Simulated “Visit” labels based on engagement thresholds

• Trained a Random Forest Classifier to predict visit likelihood

• Evaluated feature importance (e.g. website visits, discounts, referrals)

• Built a visual dashboard + confusion matrix

• Delivered strategic insights + a case study to guide retargeting

​

Key Insight:​ Website visits, social proof, and value-seeking behaviors were top visit predictors — not influencer engagement or app usage.

​

PROJECT 2: Micro-Targeted Ad Campaign by ZIP Code

​

What I Did?

​

​I developed a ZIP-code-based advertising strategy using AI to generate localized ad copy and creative prompts tailored to different Atlanta communities.

Why I Did It?

​

​Generic messaging doesn’t work for everyone. NCCHR’s mission touches different communities in different ways — from educators to activists to artists. This campaign bridges that gap with message-persona alignment.

Recommended Action Plan Based on Action Plan:

How I Did It?

​

• Simulated 5 ZIP-code personas using neighborhood-level demographic and cultural data

• For each ZIP: mapped tone, community interest, and call-to-action

• Used GPT-style logic to auto-generate custom Facebook/Instagram ads

• Created visual prompts for designers or AI tools (e.g. Midjourney, DALL·E)

• Packaged all ads into a campaign-ready Excel sheet + strategy doc

Sample Ad Copy:

“Explore Civil Rights History in Downtown ATL. Young Professionals — this is your moment to connect with the stories that shaped us. Free with Student ID.”

Visual Prompt Example:

​

 “A bold ad showing young professionals engaging with civil rights exhibits in a modern Atlanta cityscape.”

​

Impact & Takeaways:

​

• Combined behavioral prediction + hyper-local targeting = conversion at scale

• Aligned deeply with NCCHR’s mission: equity, education, and engagement

• Created a repeatable AI framework for other nonprofits or cultural institutions

​

Tools Used:

​

• Python (Pandas, Scikit-learn, Seaborn)

• ChatGPT (prompt design + ad copy generation)

• Excel, Meta Ads formatting

• Plotly, Word, Notion (for dashboard, docs, and campaign assets)