How dudi0d/project_work works
Project type and purpose
• This is a standalone desktop ML application written in Python.
• It consists of a PyQt5-based GUI front end (“pyqt_file.py”), two data-processing scripts (“big_dataset.py” and “pressure_dataset_2.csv” + other CSVs), and neural-network models (with GRU layers) for forecasting oil and gas data.
File structure
• Top-level scripts:
– pyqt_file.py → PyQt5 GUI, drives the inference pipeline.
– big_dataset.py → Batch/large-dataset evaluation of the same models.
• Data folder (implicit via CSV files):
– pressure_dataset_2.csv, “Скважина 20332…csv” etc → raw measurement time-series.
• README.md → high-level description, example screenshots, links to GRU model usage.
README for additional insights
• The GUI allows interactive prognostics using pre-trained NNs.
• big_dataset.py tests model compatibility with larger datasets (batch mode).
• Underlying ML framework isn’t named, but GRU layers imply PyTorch or TensorFlow/Keras.
Main components & interactions
a. Components
1. GUI Layer (PyQt5 front end)
2. Data Loader / Preprocessor (CSV reader, pandas/numpy routines)
3. Neural-Network Module (GRU-based forecasting engine)
4. Dataset Storage (CSV files on disk)
5. Batch-Processing Script (big_dataset.py)
6. Visualization & Output (plots, tables in the GUI)
b. Relationships & Data Flow
• User → GUI → chooses CSV file → Data Loader → Neural-Network → results → GUI → display.
• big_dataset.py runs offline: CSV set → Data Loader → Neural-Network → evaluation report (no GUI).
c. Architectural Patterns & Principles
• Modular separation of concerns: GUI vs. data vs. model logic.
• Single-process, event-driven GUI for interactive inference.
• Script-driven batch pipeline for large-scale testing.
d. Technologies & Frameworks
• Python 3.x
• PyQt5 for GUI
• pandas / numpy for data handling
• PyTorch or TensorFlow/Keras for GRU layers
• CSV storage