Comparison of Machine Learning Algorithms in Oil Production Prediction

Introduction00:00

The video opens with a greeting and introduces the topic of comparing machine learning algorithms for predicting oil production. It highlights the significance of machine learning in improving economic evaluations in the oil industry.

Research Methodology05:07

The methodology section outlines the steps for comparing algorithms, including data preparation, model building, and evaluation. Data normalization and outlier detection are key processes discussed.

Experiment 1: Single-output Models10:24

This section examines classical machine learning models, such as Random Forest and Support Vector Regression, for predicting well pressure. Support Vector Regression emerges as the most effective model.

Experiment 2: Multi-output Models15:42

Focus is on models predicting both well pressure and oil flow. The models show a good fit in training data but have larger discrepancies in test data, with Support Vector Regression again performing the best.

Experiment 3: Deep Learning Models20:51

Deep learning models including Convolutional Neural Networks and Long Short-Term Memory are tested. Results show good overall fit but varying performance, struggling with sudden changes in the data.

Discussion & Conclusion25:34

The video discusses the pros and cons of different algorithms, emphasizing the computational efficiency and effectiveness of classical models like Support Vector Regression in this context.