📱 Ideal Product Detection — Flagship Smartphone Analysis

An end-to-end Data Science project to detect the most ideal flagship smartphones (2020–2022) using ML price prediction, user survey analysis, and interactive Tableau dashboards — for both Gaming and Non-Gaming categories.

🔗 View Live Tableau Dashboard 📋 View Original Survey Form

---

🎯 Objective

There is no single smartphone that has all ideal features under a reasonable price. This project aims to:

• Predict the ideal price of a flagship phone based on user-preferred specs
• Identify real existing devices that come closest to those ideal specs
• Compare ideal vs real devices using interactive Tableau dashboards

---

🗂️ Project Pipeline

Data Collection → Feature Engineering → ML Model → Streamlit Web App
      ↓                                                      ↓
 Survey (Google Form)  →  Ideal Spec Analysis  →  Tableau Dashboard

---

📦 Repository Structure

├── FLAGSHIP 20-22 Data.xlsx                          # Raw collected dataset (69 devices)
├── Ideal_Product_Detection_Feature_Engineering_      # Feature engineering notebook
│   and_ML_Model.ipynb
├── Ideal_spec_finding_analysis(from_survey_data).ipynb  # Survey analysis notebook
├── code_made_by_sreehari.ipynb                       # Custom countplot utility function
├── flaghship pycharm.txt                             # Streamlit web app source code
├── Best_device_Tblx_dashboradlink.txt                # Tableau dashboard link
├── Ideal Product detection-Report.pdf                # Full project report
└── Project_Summary.txt                               # Project overview

---

📊 Dataset

• Source: Gadgets360, 91mobiles, GSMArena
• Scope: 69 flagship smartphones from 2020–Jan 2022
• Brands: Apple, Samsung, OnePlus, Xiaomi, Poco, Realme, Vivo, Asus, IQOO, Lenovo, Motorola, Oppo
• Criteria: Snapdragon 855+, Dimensity 1000+, or Apple A-series chipset; 5G enabled
• Fields: 21 raw features → expanded to 30+ after feature engineering

---

🔧 Phase 1 — Feature Engineering

Notebook: Ideal_Product_Detection_Feature_Engineering_and_ML_Model.ipynb

• Removed unit strings (MP, GB, Kg) from columns for numeric conversion
• Split multi-value columns (e.g. 48+16+8 camera → 3 separate columns)
• Engineered PPI (Pixels Per Inch) from resolution + screen size
• Verified splits by checking unique values; corrected misaligned columns
• Final cleaned data exported as Pythonsplit FLAGSHIP 20-22 Data.csv

---

🤖 Phase 2 — ML Model (Price Prediction)

Model: Voting Regressor (RandomForest + GradientBoosting + ExtraTrees)

Metric	Score
R² Score	0.7747
MAE	0.1628 (log scale)

• Correlation analysis to select best features
• Label encoding + OneHotEncoder via ColumnTransformer
• 80/20 train-test split
• Model exported using pickle → pipe.pkl, vd.pkl

---

🌐 Phase 3 — Streamlit Web App

A web-based price predictor built with Streamlit in PyCharm.

• Select brand, chipset, display, cameras, battery, and more
• Click "Predict Product Price" → get predicted INR price instantly
• Used to predict ideal device prices for each brand with survey-identified specs

streamlit run ipd.py

---

📋 Phase 4 — Survey & Ideal Spec Analysis

Notebook: Ideal_spec_finding_analysis(from_survey_data).ipynb

• Survey created via Google Form for both gamers and non-gamers
• Responses analyzed using custom cp() countplot function
• Most-selected features identified separately for Gaming and Normal flagship categories
• Example finding: Most users preferred 480Hz touch sample rate for gaming devices, Samsung sensor for camera

---

📈 Phase 5 — Tableau Dashboard

🔗 View Interactive Dashboard

• 15 feature sheets created for gaming + normal flagship comparison
• Devices with 8+ out of 15 matching ideal features selected as best real products
• Detected: 5 best Gaming Flagships (ASUS ROG 5, Lenovo Legion Duel 2, OnePlus 9RT, Realme GT 2, GT Neo 2) and 11 best Normal Flagships
• Interactive comparison: click any brand → see how it compares to the ideal predicted device

---

🛠️ Tech Stack

Tool	Purpose
Python	Core language
Pandas, NumPy	Data manipulation
Matplotlib, Seaborn	Visualization
Scikit-learn	ML models, encoding, evaluation
Streamlit	Web app
Pickle	Model export
Tableau Public	Interactive dashboards
Google Forms	Survey
Google Colab	Notebook environment

---

▶️ How to Run

# Clone the repo
git clone https://github.com/Sreeharipavithran/PG-Diploma-Project-2022.git

# Install dependencies
pip install streamlit scikit-learn pandas numpy matplotlib seaborn openpyxl

# Run the web app (requires pipe.pkl and vd.pkl)
streamlit run "flaghship pycharm.txt"

⚠️ Note: The pre-trained model files (pipe.pkl, vd.pkl) are not included in this repo as they were built on a local system. To reproduce them, run the Feature Engineering and ML Model notebook end-to-end first, which will generate the pickle files needed for the Streamlit app.

---

📬 Connect with Me