📸 Cracking Instagram Influence: Anomaly Detection & Growth Forecasting

🚀 Project Overview

In the ever-evolving world of social media, influence is currency. But how do we differentiate genuine engagement from manipulation? How can brands predict influencer growth and audience loyalty?

This project leverages SQL, data science, and machine learning to analyze Instagram influencer behavior, detect anomalies (e.g., fake engagement, bot activity), and forecast influence growth trends—critical insights for e-commerce brands, advertisers, and social media platforms like Meta (Instagram & Facebook).

---

🎯 Key Objectives

✅ Anomaly Detection: Identifying fake engagement, bot-driven activity, and suspicious patterns
✅ Growth & Engagement Forecasting: Predicting follower trends and influencer impact
✅ Market Insights: Evaluating the ROI of influencer partnerships
✅ Data-Driven Decision Making: Providing actionable insights for brand collaborations

---

📊 Dataset Overview

This study analyzes top Instagram influencers, extracting:

• Follower Growth: Tracking audience trends
• Engagement Metrics: Likes, comments, shares, and total influence score
• Anomaly Indicators: Unusual spikes, bot-driven engagement, and sudden audience drops
• Influencer Classification: Categorizing influencers based on reach & authenticity

---

🔍 SQL-Driven Insights

This project starts with SQL queries to extract, clean, and analyze data.

✅ Example Query: Detecting Influencers with Suspicious Engagement Rates

SELECT influencer_name, followers, engagement_rate, avg_likes, total_likes
FROM instagram_data
WHERE engagement_rate > 50 OR engagement_rate < 1
ORDER BY engagement_rate DESC;

💡 Findings:
✔ Some influencers have engagement rates above 50%—a potential red flag for fake activity.
✔ Accounts with massive followers but low engagement indicate bot-driven follower inflation.

---

🚀 Anomaly Detection: Spotting Fake Engagement & Bot Activity

💡 Why is Anomaly Detection Important?
Instagram is flooded with fake followers and bot-driven likes—distorting marketing investments.
This study applies unsupervised learning techniques to flag suspicious influencer behavior.

📌 Key Anomalies Detected:

📍 Fake Engagement: Unnaturally high like-to-follower ratios suggest purchased likes.
📍 Bot Activity: Accounts with low comment diversity but high engagement.
📍 Follower-Engagement Mismatch: High followers but low average likes = ghost followers.
📍 Rapid Engagement Drops: Sudden audience decline → Possible shadowban or lost credibility.

✅ Example Approach: Isolation Forest for Anomaly Detection

from sklearn.ensemble import IsolationForest

model = IsolationForest(contamination=0.05)  # Flagging top 5% anomalies
df["anomaly_score"] = model.fit_predict(df[["followers", "engagement_rate", "avg_likes"]])
df[df["anomaly_score"] == -1]  # Outlier detection

🔹 Result: Accounts with artificially inflated engagement flagged as anomalies.

---

📈 Influence Growth & Engagement Prediction

💡 Why Predict Influencer Growth?
Brands & marketers invest heavily in influencers—but how do they know who will remain relevant?
This study builds a time-series model to predict influencer growth trends.

✅ Example Approach: Facebook Prophet for Time-Series Forecasting

from fbprophet import Prophet

model = Prophet()
df_train = df[['date', 'followers']].rename(columns={'date': 'ds', 'followers': 'y'})
model.fit(df_train)
future = model.make_future_dataframe(periods=90)  # Predicting next 3 months
forecast = model.predict(future)

🔹 Result: Predicts audience growth for data-driven marketing decisions.

✅ Key Takeaways:
✔ Verified accounts have more stable engagement trends.
✔ High engagement volatility suggests possible audience manipulation.
✔ Consistent engagement growth → More reliable influencers for brand partnerships.

---

📊 Data Visualization & Insights

📌 Heatmaps – Analyzing correlation between followers, engagement, and authenticity
📌 Trend Analysis – Forecasting follower & engagement growth
📌 Cluster Analysis – Identifying different influencer types (nano, micro, macro, mega)

✅ Example: Engagement vs. Followers Scatter Plot

import seaborn as sns
import matplotlib.pyplot as plt

sns.scatterplot(x=df["followers"], y=df["engagement_rate"], hue=df["anomaly_score"])
plt.title("Engagement vs. Followers (Anomaly Detection)")
plt.show()

---

🎯 Why This Project Stands Out for Meta's Data Science Team

This project aligns with Meta’s Data Scientist role by demonstrating:
✔ Advanced SQL & Python expertise for data-driven insights
✔ Product Analytics Mindset – Understanding user behavior & engagement patterns
✔ Machine Learning for Anomaly Detection & Forecasting
✔ Data-Driven Storytelling – Turning insights into strategic decisions

---

🔮 Future Scope

🔹 Real-time influencer monitoring system using Meta API
🔹 Automated fraud detection & influencer classification
🔹 Building an ETL pipeline for influencer analytics

---

🛠 How to Run This Project

1. Clone the repo:

   git clone https://github.com/shrunalisalian/instagram-influencer-analysis.git

2. Install dependencies:

   pip install -r requirements.txt

3. Run the Jupyter Notebook:

   jupyter notebook instagram-case-study.ipynb

---

📌 Connect with Me

• LinkedIn: Shrunali Salian
• Portfolio: https://portfolio-shrunali-suresh-salians-projects.vercel.app/
• Email: Your Email

---

🚀 Final Thoughts

This project goes beyond basic analytics—it provides real-world insights into Instagram’s influence economy. Anomaly detection, engagement forecasting, and SQL-driven insights make this study a powerful tool for e-commerce brands, digital marketers, and Meta’s data science team.