Project MLP, which stands for Machine Learning Portfolio, is a collection of mini machine learning projects designed to showcase my understanding of ML concepts. I created this repository to evaluate my skills in ML and my capability to solve real-world problems. Throughout this project, I gained valuable insights into plotting different types of data and applying various algorithms to address them.
-
House Price Prediction: Using simple and multiple linear regression.
-
China's GDP Prediction: Utilizing non-linear regression (sigmoid function).
-
Fish Weight Prediction: Implementing simple and multiple polynomial regression.
-
Customer Churn Detection: Applying decision trees.
-
Heart Attack Probability Prediction: Utilizing the k-nearest neighbors algorithm.
-
Wine Quality Prediction: Using logistic regression.
-
Student Performance Prediction: Applying the support vector machine algorithm.
-
Customer Segmentation: Utilizing three different algorithms:
-
Movie Recommendation System: Implementing two different algorithms:
Before running this project on your local machine, I recommend installing Anaconda, which comes with many of the necessary tools pre-installed.
-
Clone the repository:
git clone https://github.com/work-atrvoid/project-mlp.git
-
Navigate to the project folder:
cd project-mlp -
Install dependencies:
pip install -r requirements.txt
-
Open the Jupyter notebooks:
jupyter notebook
- Python programming language
- Jupyter Notebook
- NumPy library for mathematical purposes, array/matrix manipulations, and more
- Pandas library for DataFrame and dataset manipulations
- Scikit-learn library for machine learning models, preprocessing datasets, and model evaluation
- Matplotlib library for plotting and visualization
- Seaborn library for styling and creating pair plots
- Plotly library for interactive plots
- Dtreeviz library for visualizing decision trees
- SciPy library for optimization and dendrogram clustering
-
The application of various machine learning algorithms and their suitability for different types of problems.
-
Data preprocessing techniques and their impact on model performance, such as one-hot encoding, label encoding, train-test splitting, normalizing with different methods (e.g., standardization or min-max scaling), and more.
-
The importance of data visualization in understanding model outputs and improving insights.
-
The need for careful analysis and consideration. For example, in GDP prediction, while a growth rate plot might suggest a polynomial relationship, it's important to recognize that moving from 0 to 0.9 is relatively easy, whereas moving from 0.9 to 1 is more challenging. Developing countries can progress more easily by following the paths of developed countries, while developed countries must innovate to achieve further growth, which led me to use a sigmoid non-linear function instead of a polynomial one.
-
Sometimes, finding that there is no relationship between two parameters can be as insightful as discovering a relationship. For instance, in student performance prediction, understanding that there may not be a straightforward relationship between weekly study hours and better grades led me to consider the possibility of a hyperplane in a higher-dimensional space, which ultimately helped in solving the problem using the support vector machine algorithm.
-
Hands-on experience with multiple libraries and tools commonly used in the machine learning field.
I plan to explore deep learning and additional classic ML algorithms. If I discover any new mini-projects that align with this portfolio, I will gladly include them!
If you have any questions or feedback regarding this project, please feel free to contact me. I welcome any thoughts or suggestions you may have!
If you enjoyed my work or found this portfolio helpful, please consider giving it a star! Your support encourages me to keep improving and sharing my projects with the community. Thank you!