CAR CLASSIFICATION PROJECT

K-NEAREST NEIGHBOURS

a classification algorithm / deals with irregular data

Attributes

1. Buying cost _"buying"
2. Maintenance _"maint"
3. Number of doors _"door"
4. Number of Persons _"persons"
5. Boot size _"lug_boot"
6. Safety degree _"safety"

Label / The prediction

7. class = ["unacc", "acc", "good", "vgood"]

Requirements

1. pandas

   python import pandas as pd

2. Numpy

   python import numpy as np

3. sklearn

   import sklearn 
   from sklearn import linear_model, preprocessing
   from sklearn.utils import shuffle 
   from sklearn.neighbors import KNeighborsClassifier

Steps of the project:

STEP 1:

• We have to read in our dataset. Using panda.

  data = pd.read_csv("car.data")

STEP 2:

• To encode the non-integral data values

  encode = preprocessing.LabelEncoder()
  
  buying = encode.fit_transform(list(data["buying"]))
  maint = encode.fit_transform(list(data["maint"]))
  door = encode.fit_transform(list(data["door"]))
  persons = encode.fit_transform(list(data["persons"]))
  lug_boot = encode.fit_transform(list(data["lug_boot"]))
  safety = encode.fit_transform(list(data["safety"]))
  cls = encode.fit_transform(list(data["class"]))

STEP 3:

• Here we define what we want to predict. The label.

  predict = "cls"

  x = list(zip(buying, maint, door, persons, lug_boot, safety)) 

• This line defines attributes that will help with prediction.

  y = list(cls)

• This line gives only the 'class' value.

STEP 4:

• We divide x & y into four('x train', 'y train', 'x test', 'y test'). Using sklearn!

  x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(x, y, test_size=0.1) 

• This line splits our data X and Y, This is for TRAINING & TESTING.

• The line 'test_size=0.1', means that from our dataset 10% will be used for testing.

STEP 5:

• create the training model.

  model = KNeighborsClassifier(n_neighbors=7)
  model.fit(x_train, y_train)

• n_neighbours=7 , means a maximum of 7 neighbours is allowed

• You can use more or less eg: n_neighbours=5, n_neighbours=11 etc..

• only odd numbers are needed.

  accuracy = model.score(x_test, y_test)
  print(accuracy)

• model-score, finds how accurate the model is!

HOW K-NEAREST NEIGHBOURS works!!!

STEP 6:

```python 
prediction = model.predict(x_test)
names = ["unacc", "acc", "good", "vgood"]

for x in range(len(prediction)):
    print(names[prediction[x]], x_test[x], names[y_test[x]])
    # or
    print("Predicted: ", names[prediction[x]], "Actual: ", names[y_test[x]])
    
```

• On the first line our model makes a prediction.
• A for loop to iterate through each prediction!
• names variable & n helps give proper class names to our output.

STEP 7:

• Saving our model

• If we were to save our model it would consume alot of space, it's one limitations of this algorithm.

• This because of the magnitude calculations of each single value with all the other possibilities.

The END