Skip to content
Snippets Groups Projects
Commit e867a84d authored by Daniel Müller's avatar Daniel Müller :speech_balloon:
Browse files

Remove empty cell SVM solution

parent 4089f4c8
No related branches found
No related tags found
1 merge request!124Fix CNN + Algorithm Comparison
Hide whitespace changes
Inline Side-by-side
notebooks/exercises/solution_support_vector_machine.ipynb
+ 2
− 10
View file @ e867a84d
%% Cell type:markdown id:4faad863-6d6b-4c19-b9a5-67131aa6f123 tags: %% Cell type:markdown id:4faad863-6d6b-4c19-b9a5-67131aa6f123 tags:
# Support Vector Machine # Support Vector Machine
In this exercise we will see another more different usage of the SVM. Maybe you already saw the problem of handwritten-digit-classification. Its a common problem for neural networks and we will now try to solve it by using a mathematical model. In this exercise we will see another more different usage of the SVM. Maybe you already saw the problem of handwritten-digit-classification. Its a common problem for neural networks and we will now try to solve it by using a mathematical model.
Remember there is no right and wrong solution to this. The solution-notebook just shows you an example. Remember there is no right and wrong solution to this. The solution-notebook just shows you an example.
## Imports ## Imports
These are the imports we will use. These are the imports we will use.
%% Cell type:code id:6442efc2-de82-418c-8de3-f1ad68b506c9 tags: %% Cell type:code id:6442efc2-de82-418c-8de3-f1ad68b506c9 tags:
``` python ``` python
import numpy as np import numpy as np
import pandas as pd import pandas as pd
from sklearn.model_selection import train_test_split from sklearn.model_selection import train_test_split
from sklearn.svm import SVC from sklearn.svm import SVC
from sklearn.metrics import confusion_matrix from sklearn.metrics import confusion_matrix
from sklearn.preprocessing import scale from sklearn.preprocessing import scale
from sklearn.model_selection import validation_curve from sklearn.model_selection import validation_curve
from sklearn.model_selection import KFold from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score from sklearn.model_selection import cross_val_score
from sklearn.model_selection import GridSearchCV from sklearn.model_selection import GridSearchCV
from sklearn import metrics from sklearn import metrics
from sklearn.metrics import confusion_matrix from sklearn.metrics import confusion_matrix
import matplotlib.pyplot as plt import matplotlib.pyplot as plt
import seaborn as sns import seaborn as sns
import warnings import warnings
warnings.filterwarnings('ignore') warnings.filterwarnings('ignore')
``` ```
%% Cell type:markdown id:021ba92c-bf02-4001-bd8f-3b78bd7ee5dd tags: %% Cell type:markdown id:021ba92c-bf02-4001-bd8f-3b78bd7ee5dd tags:
We import the data. We import the data.
%% Cell type:code id:65b6375e-abe9-4af5-9cfc-335ffb539fe1 tags: %% Cell type:code id:65b6375e-abe9-4af5-9cfc-335ffb539fe1 tags:
``` python ``` python
train_data = pd.read_csv("../data/Digits/train_digits.csv.gz", nrows=20000) train_data = pd.read_csv("../data/Digits/train_digits.csv.gz", nrows=20000)
test_data = pd.read_csv("../data/Digits/test_digits.csv.gz",nrows=5000) test_data = pd.read_csv("../data/Digits/test_digits.csv.gz",nrows=5000)
#train_data = train_data[:20000] #train_data = train_data[:20000]
#test_data = test_data[:10000] #test_data = test_data[:10000]
``` ```
%% Cell type:markdown id:703956d1-0290-411c-92d1-d6d5cb844a1f tags: %% Cell type:markdown id:703956d1-0290-411c-92d1-d6d5cb844a1f tags:
## Task 1: ## Task 1:
Visualize the data. Pyplot and seaborn is imported. This task is only for the purpose to understand the struture of the data. Visualize the data. Pyplot and seaborn is imported. This task is only for the purpose to understand the struture of the data.
%% Cell type:code id:b70150bb-d9b3-443b-ac30-d739b1f82973 tags: %% Cell type:code id:b70150bb-d9b3-443b-ac30-d739b1f82973 tags:
``` python ``` python
# Visualizing the number of class and counts in the datasets # Visualizing the number of class and counts in the datasets
# The head() function is also really usefull # The head() function is also really usefull
plt.plot(figure = (16,10)) plt.plot(figure = (16,10))
g = sns.countplot(train_data["label"], palette = 'icefire') g = sns.countplot(train_data["label"], palette = 'icefire')
plt.title('Number of digit classes') plt.title('Number of digit classes')
train_data.label.astype('category').value_counts() train_data.label.astype('category').value_counts()
plt.show() plt.show()
``` ```
%% Output %% Output
%% Cell type:markdown id:e6bad466-e596-4c45-854f-7700c63c057e tags: %% Cell type:markdown id:e6bad466-e596-4c45-854f-7700c63c057e tags:
## Task 2: Preprocessing ## Task 2: Preprocessing
At the moment the labels are within the actual data with the 'label' tag. And also we need to use scale() on the rest of the data. At the moment the labels are within the actual data with the 'label' tag. And also we need to use scale() on the rest of the data.
You now need to prepocess the data. Also think about what the SVM needs and what we have in our data and do we actually need to do anything to it. You now need to prepocess the data. Also think about what the SVM needs and what we have in our data and do we actually need to do anything to it.
Finally we will use the train_text_split function to make a test data set. Finally we will use the train_text_split function to make a test data set.
%% Cell type:code id:beb2c23f-f90c-44dd-8467-a57c3a65399c tags: %% Cell type:code id:beb2c23f-f90c-44dd-8467-a57c3a65399c tags:
``` python ``` python
# Separating the X and Y variable # Separating the X and Y variable
y = train_data['label'] y = train_data['label']
# Dropping the variable 'label' from X variable # Dropping the variable 'label' from X variable
X = train_data.drop(columns = 'label') X = train_data.drop(columns = 'label')
X_scaled = scale(X) X_scaled = scale(X)
# train test split # train test split
X_train, X_test, y_train, y_test = train_test_split(X_scaled, y, test_size = 0.3, train_size = 0.2 ,random_state = 10) X_train, X_test, y_train, y_test = train_test_split(X_scaled, y, test_size = 0.3, train_size = 0.2 ,random_state = 10)
``` ```
%% Cell type:markdown id:93f99db2-cfde-4763-b444-e2cfc9a25ded tags: %% Cell type:markdown id:93f99db2-cfde-4763-b444-e2cfc9a25ded tags:
## Task 3: Model training ## Task 3: Model training
The third and final task is to actually load the model, train it and finally test its accuracy. The third and final task is to actually load the model, train it and finally test its accuracy.
You already know the method from the actual notebook of the SVM. You already know the method from the actual notebook of the SVM.
Finally we will visualize the results. In this example it will be best to use a confusion matrix. Finally we will visualize the results. In this example it will be best to use a confusion matrix.
%% Cell type:code id:8546673f-3319-4ec0-9090-fa75a652e0ab tags: %% Cell type:code id:8546673f-3319-4ec0-9090-fa75a652e0ab tags:
``` python ``` python
model_linear = SVC(kernel='linear') model_linear = SVC(kernel='linear')
model_linear.fit(X_train, y_train) model_linear.fit(X_train, y_train)
# predict # predict
y_pred = model_linear.predict(X_test) y_pred = model_linear.predict(X_test)
``` ```
%% Cell type:code id:1013637b-250d-400b-95ef-cee8a39cff15 tags: %% Cell type:code id:1013637b-250d-400b-95ef-cee8a39cff15 tags:
``` python ``` python
# accuracy # accuracy
print("accuracy:", metrics.accuracy_score(y_true=y_test, y_pred=y_pred), "\n") print("accuracy:", metrics.accuracy_score(y_true=y_test, y_pred=y_pred), "\n")
# cm # cm
mat = metrics.confusion_matrix(y_true=y_test, y_pred=y_pred) mat = metrics.confusion_matrix(y_true=y_test, y_pred=y_pred)
plt.figure(figsize=(11,11)) plt.figure(figsize=(11,11))
sns.heatmap(mat, annot=True, cmap='Blues',fmt='d', cbar=False, square=True) sns.heatmap(mat, annot=True, cmap='Blues',fmt='d', cbar=False, square=True)
plt.xlabel('true label') plt.xlabel('true label')
plt.ylabel('predicted label'); plt.ylabel('predicted label');
``` ```
%% Output %% Output
accuracy: 0.9093333333333333 accuracy: 0.9093333333333333
%% Cell type:code id:9cdb44b3-956c-4e52-aa09-e9e263d9f0a6 tags:
``` python
```
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment