🧠 tml-visitor-behavior-analysis

A case study in data preprocessing and behavioral analysis of real visitor interactions from TechnoMagicLand (TML).
The project leverages clustering, correlation, and statistical visualization in R to uncover key usage patterns and optimize user engagement.

---

🧾 Project Scope

This project follows a full CRISP-DM cycle for behavior-driven analytics based on museum interaction logs.

🧪 Key Objectives

• Clean and prepare real-world visitor data
• Identify high-value vs casual users
• Understand exhibit engagement patterns
• Cluster users based on time, visits, and interaction profiles
• Visualize behavioral and statistical relationships

🧮 Code execution is available in the R script:
➡️ all-tables-preprocessed_final_1.R

---

🧰 Tools & Technologies

• Language: R
• Libraries: dplyr, ggplot2, tidyr, cluster, factoextra, corrplot
• Clustering: k-means with elbow method
• Visualization: Histograms, scatter plots, boxplots, heatmaps
• Data source: Interaction logs from TechnoMagicLand

---

📊 Visual Outputs

The analysis includes a variety of visualizations available in the repository under the PNG files section. These include:

• Boxplots showing the distribution of total points and interaction durations
• Histograms for visitor scores, durations, and behavioral segments
• Bar charts ranking exhibits by average time, total interactions, and difficulty
• Correlation heatmaps for both exhibit co-occurrence and user metrics
• K-Means clustering outputs with elbow method validation and scatter plots
• Visitor segmentation visuals (e.g. top users by points, visits, or consistency)

All visualizations can be found in the /figures or root folder as .png files with descriptive filenames.

---

🔍 Key Findings

• 🧍‍♂️ Most visitors only visit once, but some accumulate many points
• 🧠 High repeat users tend to interact with more exhibits on average
• 🎯 Certain exhibits have consistently high interaction time but low scores, indicating high difficulty or engagement
• 📉 Many visitors drop off after only 1–2 interactions
• 🤖 Clustering identified meaningful segments (e.g. speedrunners vs explorers)

---

📁 Repository Structure

📦 tml-visitor-behavior-analysis ├── 📊 PNG visualizations ├── 📄 all-tables-preprocessed_final_1.R ├── 📘 README.md ├── 📑 LICENSE (MIT) ├── 📕 Report PDF (Technical Summary)

---

🧠 Topics

r, clustering, data-preprocessing, data-visualization,
visitor-behavior, user-segmentation, education-project, technomagicland

---

📜 License

This project is licensed under the MIT License – see LICENSE for details.

---

🙌 Credits

Conducted as part of a data science and education analytics project based on real museum interaction data.
Developed by Sergey Filipov