
import pandas as pd
import numpy as np
from matplotlib import pyplot as plt
plt.figure(figsize=(12, 6))

#Create your df here:
df = pd.read_csv("capstone_starter/profiles.csv")
df.head()

<Figure size 864x432 with 0 Axes>


df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 59946 entries, 0 to 59945
Data columns (total 31 columns):
 #   Column       Non-Null Count  Dtype  
---  ------       --------------  -----  
 0   age          59946 non-null  int64  
 1   body_type    54650 non-null  object 
 2   diet         35551 non-null  object 
 3   drinks       56961 non-null  object 
 4   drugs        45866 non-null  object 
 5   education    53318 non-null  object 
 6   essay0       54458 non-null  object 
 7   essay1       52374 non-null  object 
 8   essay2       50308 non-null  object 
 9   essay3       48470 non-null  object 
 10  essay4       49409 non-null  object 
 11  essay5       49096 non-null  object 
 12  essay6       46175 non-null  object 
 13  essay7       47495 non-null  object 
 14  essay8       40721 non-null  object 
 15  essay9       47343 non-null  object 
 16  ethnicity    54266 non-null  object 
 17  height       59943 non-null  float64
 18  income       59946 non-null  int64  
 19  job          51748 non-null  object 
 20  last_online  59946 non-null  object 
 21  location     59946 non-null  object 
 22  offspring    24385 non-null  object 
 23  orientation  59946 non-null  object 
 24  pets         40025 non-null  object 
 25  religion     39720 non-null  object 
 26  sex          59946 non-null  object 
 27  sign         48890 non-null  object 
 28  smokes       54434 non-null  object 
 29  speaks       59896 non-null  object 
 30  status       59946 non-null  object 
dtypes: float64(1), int64(2), object(28)
memory usage: 14.2+ MB


# Let's look at the "age" column ...

plt.hist(df.age, bins=20)
plt.xlabel('age')
plt.ylabel('frequency')
plt.xlim(16, 80)

(16.0, 80.0)


# Let's dig a little bit into the 'job' column ...

df.job.head()

0                 transportation
1           hospitality / travel
2                            NaN
3                        student
4    artistic / musical / writer
Name: job, dtype: object


# We see that at leaast one of the rows has a NaN in it, so we'll check to see how many we have in the entire dataframe ...
nan_count = df['job'].isna().sum()
nan_count

8198


# let's drop those rows and check the dataframe again after handling the missing data ...

df = df.dropna(subset=['job'])
nan_count = df['job'].isna().sum()
nan_count

0


import seaborn as sns
#sns.set(rc={'figure.figsize':(30,25)})
#sns.set(font_scale=1)

# Plot pairplot
sns.pairplot(df)

# Show the plot
plt.show()


sns.countplot(x='job', hue='orientation', data=df)
plt.xticks(rotation=90)

(array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
        17, 18, 19, 20]),
 [Text(0, 0, 'transportation'),
  Text(1, 0, 'hospitality / travel'),
  Text(2, 0, 'student'),
  Text(3, 0, 'artistic / musical / writer'),
  Text(4, 0, 'computer / hardware / software'),
  Text(5, 0, 'banking / financial / real estate'),
  Text(6, 0, 'entertainment / media'),
  Text(7, 0, 'sales / marketing / biz dev'),
  Text(8, 0, 'other'),
  Text(9, 0, 'medicine / health'),
  Text(10, 0, 'science / tech / engineering'),
  Text(11, 0, 'executive / management'),
  Text(12, 0, 'education / academia'),
  Text(13, 0, 'clerical / administrative'),
  Text(14, 0, 'construction / craftsmanship'),
  Text(15, 0, 'rather not say'),
  Text(16, 0, 'political / government'),
  Text(17, 0, 'law / legal services'),
  Text(18, 0, 'unemployed'),
  Text(19, 0, 'military'),
  Text(20, 0, 'retired')])


df.job.value_counts()

other                                7589
student                              4882
science / tech / engineering         4848
computer / hardware / software       4709
artistic / musical / writer          4439
sales / marketing / biz dev          4391
medicine / health                    3680
education / academia                 3513
executive / management               2373
banking / financial / real estate    2266
entertainment / media                2250
law / legal services                 1381
hospitality / travel                 1364
construction / craftsmanship         1021
clerical / administrative             805
political / government                708
rather not say                        436
transportation                        366
unemployed                            273
retired                               250
military                              204
Name: job, dtype: int64


plt.style.use('dark_background')
plt.tight_layout()
df['job'].value_counts().plot(kind='pie', autopct="%1.1f%%")

<AxesSubplot:ylabel='job'>


# Let's also have a look at the "orientation" column ...

df.orientation.value_counts()

straight    44657
gay          4752
bisexual     2339
Name: orientation, dtype: int64


sns.countplot(x='job', hue='orientation', data=df)
plt.xticks(rotation=90)

(array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
        17, 18, 19, 20]),
 [Text(0, 0, 'transportation'),
  Text(1, 0, 'hospitality / travel'),
  Text(2, 0, 'student'),
  Text(3, 0, 'artistic / musical / writer'),
  Text(4, 0, 'computer / hardware / software'),
  Text(5, 0, 'banking / financial / real estate'),
  Text(6, 0, 'entertainment / media'),
  Text(7, 0, 'sales / marketing / biz dev'),
  Text(8, 0, 'other'),
  Text(9, 0, 'medicine / health'),
  Text(10, 0, 'science / tech / engineering'),
  Text(11, 0, 'executive / management'),
  Text(12, 0, 'education / academia'),
  Text(13, 0, 'clerical / administrative'),
  Text(14, 0, 'construction / craftsmanship'),
  Text(15, 0, 'rather not say'),
  Text(16, 0, 'political / government'),
  Text(17, 0, 'law / legal services'),
  Text(18, 0, 'unemployed'),
  Text(19, 0, 'military'),
  Text(20, 0, 'retired')])


# Formulating a question: Can one use the data provided to predict a person's orientation? What might their "job" have to do with this, if anything ...? So, to start, let's first convert the strings in the "job" column to numbers and then replace all the strings with our numbers ...

unique_jobs = df['job'].unique()
category_to_number = {job: i for i, job in enumerate(unique_jobs)}
df['job_code'] = df['job'].replace(category_to_number)
print(df['job_code'].head())
print(df['job_code'].value_counts())

0    0
1    1
3    2
4    3
5    4
Name: job_code, dtype: int64
8     7589
2     4882
10    4848
4     4709
3     4439
7     4391
9     3680
12    3513
11    2373
5     2266
6     2250
17    1381
1     1364
14    1021
13     805
16     708
15     436
0      366
18     273
20     250
19     204
Name: job_code, dtype: int64


# And let's have a look at the values for the "sex" column too ...

df['sex'].value_counts()

m    31287
f    20461
Name: sex, dtype: int64


# Now let's also convert the "m" and "f" values in that column to numbers and then replace everything in that column with the numbers; we'll also do the same for the "orientation" column, which is the one we'll ask our model to make predictions about ...

sex_category_to_number = {
    'm': 0,
    'f': 1
}

orientation_category_to_number = {
    'straight': 0,
    'gay': 1,
    'bisexual': 2
}

df['sex_code'] = df['sex'].replace(sex_category_to_number)
df['orientation_code'] = df['orientation'].replace(sex_category_to_number)


# Next let's pull in our sklearn library and get the data normalized (docs for sklearn's MinMaxScaler is available at https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html)

import sklearn.preprocessing as preprocessing
from sklearn.preprocessing import MinMaxScaler

# We'll start by seeing if one can predict the value of the "orientation" column using the data from the converted "job" and "sex" columns:

feature_data = df[['job_code', 'sex_code']]
x = feature_data.values
min_max_scaler = preprocessing.MinMaxScaler()
x_scaled = min_max_scaler.fit_transform(x)
feature_data = pd.DataFrame(x_scaled, columns=feature_data.columns)


# After we split the dataset we'll try a K-Means Nearest Neigbors (KNN) Algorithm (docs available at https://scikit-learn.org/stable/modules/neighbors.html)

from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
from sklearn.neighbors import KNeighborsClassifier

# split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(feature_data, df['orientation_code'], test_size=0.2, random_state=42)
n_neighbors_list = range(1, 201)

best_n_neighbors = None
best_score = 0
accuracy_score = []
# loop over the values of n_neighbors and train a KNN classifier for each one
for n_neighbors in n_neighbors_list:
    # create a KNN classifier object with the current value of n_neighbors
    knn = KNeighborsClassifier(n_neighbors=n_neighbors)

    # train the classifier on the training data
    knn.fit(X_train, y_train)

    # evaluate the performance of the classifier on the test data
    y_pred = knn.predict(X_test)
    score = knn.score(X_test, y_test)
    accuracy_score.append(score)


plt.figure(figsize=(12, 6))
plt.plot(n_neighbors_list, accuracy_score)
plt.xlabel("number of neighbors (k)")
plt.ylabel("accuracy score")
plt.show()


# What about a different algorithm—say, the GaussianNB (docs at https://scikit-learn.org/stable/modules/naive_bayes.html#gaussian-naive-bayes)

from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import classification_report

# create a Gaussian Naive Bayes classifier object
nb = GaussianNB()

# train the classifier on the training data
nb.fit(X_train, y_train)

# evaluate the performance of the classifier on the test data
y_pred = nb.predict(X_test)
print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

    bisexual       0.00      0.00      0.00       470
         gay       0.00      0.00      0.00       957
    straight       0.86      1.00      0.93      8923

    accuracy                           0.86     10350
   macro avg       0.29      0.33      0.31     10350
weighted avg       0.74      0.86      0.80     10350

c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))


# What about a "Support Vectors Machine" algorithm (docs at https://scikit-learn.org/stable/modules/classes.html#module-sklearn.svm)

from sklearn.svm import SVC
from sklearn.metrics import classification_report

# create an SVM classifier object with a linear kernel
svm = SVC(kernel='linear')

# train the classifier on the training data
svm.fit(X_train, y_train)

# evaluate the performance of the classifier on the test data
y_pred = svm.predict(X_test)
print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

    bisexual       0.00      0.00      0.00       470
         gay       0.00      0.00      0.00       957
    straight       0.86      1.00      0.93      8923

    accuracy                           0.86     10350
   macro avg       0.29      0.33      0.31     10350
weighted avg       0.74      0.86      0.80     10350

c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))


from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score

# create a decision tree classifier object
dtc = DecisionTreeClassifier()

# train the model on the training data
dtc.fit(X_train, y_train)

# evaluate the performance of the model on the test data
y_pred = dtc.predict(X_test)
print(classification_report(y_test, y_pred))

# create a random forest classifier object
rfc = RandomForestClassifier()

# train the model on the training data
rfc.fit(X_train, y_train)

# evaluate the performance of the model on the test data
y_pred = rfc.predict(X_test)
print(classification_report(y_test, y_pred))

c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))

              precision    recall  f1-score   support

    bisexual       0.00      0.00      0.00       470
         gay       0.00      0.00      0.00       957
    straight       0.86      1.00      0.93      8923

    accuracy                           0.86     10350
   macro avg       0.29      0.33      0.31     10350
weighted avg       0.74      0.86      0.80     10350

              precision    recall  f1-score   support

    bisexual       0.00      0.00      0.00       470
         gay       0.00      0.00      0.00       957
    straight       0.86      1.00      0.93      8923

    accuracy                           0.86     10350
   macro avg       0.29      0.33      0.31     10350
weighted avg       0.74      0.86      0.80     10350

c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
c:\Users\KSpicer\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))


from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import learning_curve

# Define the DTC model
dtc = DecisionTreeClassifier(max_depth=3)

# Generate the learning curve
train_sizes, train_scores, test_scores = learning_curve(dtc, X_train, y_train, cv=5)

# Calculate the mean and standard deviation of the training and testing scores
train_mean = np.mean(train_scores, axis=1)
train_std = np.std(train_scores, axis=1)
test_mean = np.mean(test_scores, axis=1)
test_std = np.std(test_scores, axis=1)

# Plot the learning curve
plt.plot(train_sizes, train_mean, label='Training score')
plt.plot(train_sizes, test_mean, label='Cross-validation score')

# Add error bars
plt.fill_between(train_sizes, train_mean - train_std, train_mean + train_std, alpha=0.1)
plt.fill_between(train_sizes, test_mean - test_std, test_mean + test_std, alpha=0.1)

# Add labels and legend
plt.xlabel('Training set size')
plt.ylabel('Accuracy score')
plt.title('Decision Tree Classifier Learning Curve')
plt.legend()

# Show the plot
plt.show()


# What if we added some features to our feature_data from the original dataset? Would that give us appreciably different results, still using the same KNN Classifier?

feature_data = df[['age', 'income', 'job_code', 'sex_code', ]]
x = feature_data.values
min_max_scaler = preprocessing.MinMaxScaler()
x_scaled = min_max_scaler.fit_transform(x)
feature_data = pd.DataFrame(x_scaled, columns=feature_data.columns)

# split the data into training and testing sets
#X_train, X_test, y_train, y_test = train_test_split(df[['job_code', 'sex_code']], df['orientation_code'], test_size=0.2, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(feature_data, df['orientation_code'], test_size=0.2, random_state=42)
n_neighbors_list = range(1, 201)

best_n_neighbors = None
best_score = 0
accuracy_score = []
# loop over the values of n_neighbors and train a KNN classifier for each one
for n_neighbors in n_neighbors_list:
    # create a KNN classifier object with the current value of n_neighbors
    knn = KNeighborsClassifier(n_neighbors=n_neighbors)

    # train the classifier on the training data
    knn.fit(X_train, y_train)

    # evaluate the performance of the classifier on the test data
    y_pred = knn.predict(X_test)
    score = knn.score(X_test, y_test)
    accuracy_score.append(score)
    
plt.figure(figsize=(12, 6))
plt.plot(n_neighbors_list, accuracy_score)
plt.xlabel("number of neighbors (k)")
plt.ylabel("accuracy score")
plt.show()

	age	body_type	diet	drinks	drugs	education	essay0	essay1	essay2	essay3	...	location	offspring	orientation	pets	religion	sex	sign	smokes	speaks	status
0	22	a little extra	strictly anything	socially	never	working on college/university	about me:<br />\n<br />\ni would love to think...	currently working as an international agent fo...	making people laugh.<br />\nranting about a go...	the way i look. i am a six foot half asian, ha...	...	south san francisco, california	doesn’t have kids, but might want them	straight	likes dogs and likes cats	agnosticism and very serious about it	m	gemini	sometimes	english	single
1	35	average	mostly other	often	sometimes	working on space camp	i am a chef: this is what that means.<br />\n1...	dedicating everyday to being an unbelievable b...	being silly. having ridiculous amonts of fun w...	NaN	...	oakland, california	doesn’t have kids, but might want them	straight	likes dogs and likes cats	agnosticism but not too serious about it	m	cancer	no	english (fluently), spanish (poorly), french (...	single
2	38	thin	anything	socially	NaN	graduated from masters program	i'm not ashamed of much, but writing public te...	i make nerdy software for musicians, artists, ...	improvising in different contexts. alternating...	my large jaw and large glasses are the physica...	...	san francisco, california	NaN	straight	has cats	NaN	m	pisces but it doesn’t matter	no	english, french, c++	available
3	23	thin	vegetarian	socially	NaN	working on college/university	i work in a library and go to school. . .	reading things written by old dead people	playing synthesizers and organizing books acco...	socially awkward but i do my best	...	berkeley, california	doesn’t want kids	straight	likes cats	NaN	m	pisces	no	english, german (poorly)	single
4	29	athletic	NaN	socially	never	graduated from college/university	hey how's it going? currently vague on the pro...	work work work work + play	creating imagery to look at:<br />\nhttp://bag...	i smile a lot and my inquisitive nature	...	san francisco, california	NaN	straight	likes dogs and likes cats	NaN	m	aquarius	no	english	single

Predictive Modeling on the "OK Cupid" Profiles dataset¶

Let's get all our libraries imported and ready to go ...:¶

Let's Do a Little EDA:¶

Setting up the Data for Training:¶

Splitting-Up the Data into Training and Testing Sets and Training the Model:¶