Development » CompressPdf

Description:¶

The objective is to predict the DPI (as a regression problem) and select the encoding type (as a classification problem) using various machine learning models. Several models were tested, hyperparameters were tuned, and the best-performing models were selected based on accuracy. The models selected were a Neural Network (Keras/TensorFlow) for DPI prediction and a Random Forest Classifier for encoding type selection.

Task List:¶

• Model Selection

  • Linear Regression
  • Decision Tree
  • Random Forest
  • Support Vector Machine (SVM)
  • XGBoost
  • CatBoost
  • Gradient Descent Regressor
  • Neural Networks (Keras/TensorFlow)

• Hyperparameter Tuning

  • Tune hyperparameters for each model using grid search and randomized search.

• Model Evaluation

  • Evaluate the models using validation metrics (accuracy, RMSE, k-fold cross-validation score).
  • Compare model performances and select the top-performing ones.

• DPI Prediction (Regression Task)

  • Select Neural Network (Keras/TensorFlow) for DPI prediction.
  • Evaluate the Neural Network model, which achieved an accuracy of 82%.

• Encoding Type Prediction (Classification Task)

  • Select Random Forest Classifier for encoding type selection.
  • Evaluate the Random Forest Classifier, which achieved an accuracy of 58%.

• Final Model Integration

  • Integrate the selected models (Neural Network for DPI and Random Forest for encoding type).
  • Integrate both trained models in a single model.keras file.