Skip to content

Instantly share code, notes, and snippets.

@rnowling
Created October 13, 2015 14:36
Show Gist options
  • Save rnowling/cd05281eeafe46b38158 to your computer and use it in GitHub Desktop.
Save rnowling/cd05281eeafe46b38158 to your computer and use it in GitHub Desktop.
Customer Segmentation Pipeline Prototype
"""
Copyright 2015 Ronald J. Nowling
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
"""
import sys
import numpy as np
import matplotlib.pyplot as plt
import csv
from sklearn.decomposition import PCA
from sklearn.cluster import k_means
from scipy.stats import rankdata
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.metrics import confusion_matrix
from sklearn.naive_bayes import MultinomialNB
def read_matrix(flname):
ratings = []
max_user_id = 0
max_item_id = 0
with open(flname) as fl:
reader = csv.reader(fl, delimiter="\t")
for rec in reader:
user_id = int(rec[0])
item_id = int(rec[1])
rating = float(rec[2])
ratings.append((user_id, item_id, rating))
max_user_id = max(max_user_id, user_id)
max_item_id = max(max_item_id, item_id)
features = np.zeros((max_user_id, max_item_id))
for user_id, item_id, rating in ratings:
features[user_id - 1, item_id - 1] = rating
return features
def read_items(flname):
items = []
with open(flname) as fl:
reader = csv.reader(fl, delimiter="|")
for rec in reader:
items.append(rec[1])
return items
def plot_explained_var_ratios(dirname, pca):
plt.plot(pca.explained_variance_ratio_[:20], "b.-")
plt.xlabel("Component", fontsize=16)
plt.ylabel("Explained Variance Ratio", fontsize=16)
plt.ylim([0.0, 1.0])
plt.savefig(dirname + "/pca_explained_var_ratio.pdf", DPI=200)
def plot_clusters(dirname, proj_features, features):
plt.clf()
plt.hold(True)
ks = [5, 10, 25, 50, 100, 125, 150, 200]
proj_costs = []
full_costs = []
for k in ks:
centroids, labels, inertia = k_means(proj_features, k)
proj_costs.append(inertia)
centroids, labels, inertia = k_means(features, k)
full_costs.append(inertia)
plt.plot(ks, np.array(proj_costs) / max(proj_costs), "b.-", label="Projected")
plt.plot(ks, np.array(full_costs) / max(full_costs), "r.-", label="Full")
plt.legend(loc="upper right")
plt.xlabel("Centers", fontsize=16)
plt.ylabel("Inertia", fontsize=16)
plt.savefig(dirname + "/kmeans_inertia.pdf", DPI=200)
def plot_confusion_matrix(dirname, cm, title='Confusion matrix', cmap=plt.cm.Blues):
plt.clf()
plt.imshow(cm, interpolation='nearest')
plt.title(title)
plt.colorbar()
plt.tight_layout()
plt.ylabel('True label')
plt.xlabel('Predicted label')
plt.savefig(dirname + "/confusion_matrix.pdf", DPI=200)
if __name__ == "__main__":
movielens_dir = sys.argv[1]
data_fl = movielens_dir + "/u.data"
items_fl = movielens_dir + "/u.item"
outdir = "figures"
features = read_matrix(data_fl)
print features.shape
sparsity = float(np.count_nonzero(features)) / (features.shape[0]
* features.shape[1])
print "Sparsity", sparsity
pca = PCA()
projected = pca.fit_transform(features)
plot_explained_var_ratios(outdir, pca)
n_pcs = 10
plot_clusters(outdir, projected[:, :n_pcs], features)
_, labels, _ = k_means(projected[:, :n_pcs], 25)
nb = MultinomialNB()
nb.fit(features, labels)
predicted = nb.predict(features)
cm = confusion_matrix(labels, predicted)
plot_confusion_matrix(outdir, cm)
scaler = TfidfTransformer()
scaled = scaler.fit_transform(features, labels).todense()
print scaled.shape
items = read_items(items_fl)
for i in xrange(25):
members = scaled[labels == i, :]
print i, members.shape
avg_rank = np.mean(members, axis=0)
ranks = rankdata(avg_rank)
rank_idx = np.argsort(-1.0 * ranks)
for j in rank_idx[:10]:
print "\t", items[j], avg_rank[0, j]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment