Skip to content

docs.md

Latest commit

 

History

History
96 lines (77 loc) · 3.08 KB

docs.md

File metadata and controls

96 lines (77 loc) · 3.08 KB
Raw

Mobile Device Price Classification System Documentation

Project Overview

This project implements an AI-driven system for classifying mobile device prices based on their specifications. The system uses machine learning to predict price ranges (0-3) for mobile devices, helping sellers accurately price their products.

Implementation Details

Data Processing

  • Dataset loaded from CSV files (train.csv and test.csv)
  • Null values removed using dropna()
  • Features standardized using StandardScaler
  • Data split: 70% training, 15% validation, 15% testing

Model Architecture

  • Algorithm: Logistic Regression (selected for optimal performance)
  • Hyperparameter tuning using GridSearchCV
  • Best parameters:
    • Determined through cross-validation
    • Optimized for accuracy

Model Performance Metrics

  • Final Model Accuracy on Test Data: 97%
  • Average confidence score: 0.92
  • Consistent performance across all price ranges

Feature Analysis

Key insights from exploratory data analysis:

  • RAM shows strong positive correlation with price
  • Battery power has minimal impact on pricing
  • Screen resolution (px_height, px_width) shows limited correlation with price

Algorithm Selection Rationale

Compared against other algorithms:

Algorithm Speed Accuracy Resource Usage Selected
Logistic Regression Fast High Low ✓
SVM Moderate High High
Random Forest Moderate Very High High
KNN Classifier Fast Moderate Moderate

Logistic Regression was chosen for:

  • Fast training time
  • Low resource requirements
  • High accuracy (97%)
  • Suitable for linear classification tasks

Model Persistence

  • Model saved using joblib
  • Scaler saved separately for preprocessing new data
  • File names:
    • best_model.pkl
    • standard_scaler.pkl

Dataset Description

Input Features

  • Device Specifications
    • battery_power: Total energy a battery can store (mAh)
    • blue: Bluetooth availability (1/0)
    • clock_speed: Microprocessor execution speed (GHz)
    • dual_sim: Dual SIM support (1/0)
    • fc: Front Camera megapixels
    • four_g: 4G support (1/0)
    • int_memory: Internal Memory (GB)
    • m_dep: Mobile Depth (cm)
    • mobile_wt: Device weight
    • n_cores: Processor core count
    • pc: Primary Camera megapixels
    • px_height: Pixel Resolution Height
    • px_width: Pixel Resolution Width
    • ram: Random Access Memory (MB)
    • sc_h: Screen Height (cm)
    • sc_w: Screen Width (cm)
    • talk_time: Battery life during calls
    • three_g: 3G support (1/0)
    • touch_screen: Touch screen availability (1/0)
    • wifi: WiFi support (1/0)

Target Variable

  • price_range: Device price category
    • 0: Low cost
    • 1: Medium cost
    • 2: High cost
    • 3: Very high cost

Model Performance

The current model implementation achieves:

  • Average confidence score: 0.92
  • Consistent performance across all price ranges
  • High prediction reliability