A Guide to A Beginner Machine Learning Model.

Prerequisites:

  • Basic Understanding of Python

Make sure you have the following installed on your device

  • Python (the programming language)
  • Tensorflow
  • Mathplotlib(optional)
  • Numpy(optional)

What will our Model do?

The model we will be constructing is one of the most basic models and we will be using it to predict the y coordinate of a line

First we will start with out imports

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

Next Lets define our data

Our training Data - the data our model will train on

x_points = np.array([3, 5, 7, 9])

y_points = np.array([6, 10, 14, 18])

Our Test Data - this is the x axis we will be predicting

test_x = np.array([2, 4, 7, 20])

Building our Model

First we have to define our model - for this blog we will be using a premade model for simplicity

model = tf.keras.Sequential([
    tf.keras.layers.Dense(units=1, input_shape=[1])
])

Compiling our Model

Next we need to build our model

Optimizer is build in so we don't need to worry about that

learn more about loss

here

 
model.compile(optimizer='sgd', loss='mean_squared_error')

Training the Model

We want to train out model on the training data so it can predict on the test data

Epochs is how many times the model trains on our data. Typically this varies for the dataset but in this case we are using 100

model.fit(x_points, y_points, epochs=100)

Testing the Model

We now have our model built now we need to test it

This is predicting on the test X data we defined earlier

test_y = model.predict(test_x)

Plotting Our Data

Now that we have our predicted model we can plot our data using mathplotlib

plt.plot(x_points, y_points, label='Train Data')
plt.plot(test_x, test_y, label='Test Data')
plt.legend()
plt.show()

This Graph will Show

ML Graph 
Train Data:
x [3, 5, 7, 9]
y [6, 10, 14, 18]

Test Data
x [2, 4, 7, 20]

Model Predicted
y [ 4.0757785, 8.044292, 13.997064, 39.792404 ]

Actual
y [4, 8, 14, 40]

It is important to note that machine learning does not typically have exact answers. It deals in probability and as you can see it is very close to the actual data

Recommended Course to Learn More

https://developers.google.com/machine-learning/crash-course

Edit this page on GitHub