---
aliases:
  - Applied ML, Applied Machine Learning
---


Additional resources: 
[[Python Data Science Handbook.pdf]]
[Train Test Split: What it Means and How to Use It | Built In](https://builtin.com/data-science/train-test-split)

1. Based on course: [Machine Learning, Data Science and Deep Learning with Python | Udemy](https://www.udemy.com/course/data-science-and-machine-learning-with-python-hands-on/learn/lecture/15090136#overview)
2. Install python, download [Course Materials: Machine Learning, Data Science, and Deep Learning with Python - Sundog Education with Frank Kane (sundog-education.com)](https://www.sundog-education.com/machine-learning/)
3. Create environment. 
	1. Open Anaconda shell, point to ML_Projects directory and open Jupyter Notebook 
```python
cd c:\MLCourse 
jupyter notebook
```
3. Load python packages and libraries

```python
import numpy  #supports sophisticated mathematical functions
from pylab import *  #module that bulk imports pyplot and numpy
from scipy.stats import *  #supports scientific and algebraic functions
import matplotlib.pyplot  #supports plots, charts
from scipy import *  #supports scientific functions
import pandas #supports data analysis, manipulation of numeric tables, checking missing data
```

1. Load your test, train data
    From computer file    
```python
dataframe = pandas.read_csv(r"C:\\ML_Projects\\LinearReg\\test.csv")
dataframe.head()  #to check top 5 rows of data and the headers
```
- Linear regression (also see [[Common AI ML terms]])
	- y=mx + b
	- Uses least squares method to minimize squared error between each point and the line = minimizes variance
	- Measure : error with r-squared = fraction of total variation captured by the model (0 to 1, 1 = perfect fit, 0 = none of the variance is captured)
1. Calculate slope, intercept
```python
slope, intercept, r_value, p_value, std_err = stats.linregress(X, Y)
r_value ** 2
```
2. Plot regression line
```python
def predict(X): 
return slope * X + intercept 
fitLine = predict(X) 
plt.scatter(X, Y) 
plt.plot(X, fitLine, c='r') 
plt.show()
```

- Polynomial regression - Fit data into a curve expressed by polynomial equation
	- Second order y = ax2 + bx + c
	- Using higher order may cause overfitting.
	- Measure : error with r-squared but might not be accurate if overfitting training data
- Multiple regressions
    E.g. predict car price based on body style, brand, mileage
    price of car = a + b1 x “mileage” + b2 x “no. of cylinders” + b3 x “no. of doors” — Note we are avoiding ordinal data like “brand”, “model” here that won’t work with mixing numerical data in regressions.
    Features: Mileage, brand, doors **need to be normalized first**.
    Eliminate features that don’t matter (when b1, b2 approaches zero).
    Measure : Fit with r-squared
    Assumption : Features are assumed to be independent of each other.
    *regressions don't work well with ordinal values, unless you can convert them into some numerical order that makes sense somehow.