Welcome to the third project of the Machine Learning Engineer Nanodegree! In this notebook, some template code has already been provided for you, and it will be your job to implement the additional functionality necessary to successfully complete this project. Sections that begin with 'Implementation' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. In addition, Markdown cells can be edited by typically double-clicking the cell to enter edit mode.
In this project, you will analyze a dataset containing data on various customers' annual spending amounts (reported in monetary units) of diverse product categories for internal structure. One goal of this project is to best describe the variation in the different types of customers that a wholesale distributor interacts with. Doing so would equip the distributor with insight into how to best structure their delivery service to meet the needs of each customer.
The dataset for this project can be found on the UCI Machine Learning Repository. For the purposes of this project, the features 'Channel' and 'Region' will be excluded in the analysis — with focus instead on the six product categories recorded for customers.
Run the code block below to load the wholesale customers dataset, along with a few of the necessary Python libraries required for this project. You will know the dataset loaded successfully if the size of the dataset is reported.
In [1]:
# Import libraries necessary for this project
import numpy as np
import pandas as pd
import renders as rs
import seaborn as sns
from IPython.display import display # Allows the use of display() for DataFrames
# Show matplotlib plots inline (nicely formatted in the notebook)
%matplotlib inline
# Load the wholesale customers dataset
try:
data = pd.read_csv("customers.csv")
data.drop(['Region', 'Channel'], axis = 1, inplace = True)
print "Wholesale customers dataset has {} samples with {} features each.".format(*data.shape)
except:
print "Dataset could not be loaded. Is the dataset missing?"
In [2]:
data.head()
Out[2]:
In this section, you will begin exploring the data through visualizations and code to understand how each feature is related to the others. You will observe a statistical description of the dataset, consider the relevance of each feature, and select a few sample data points from the dataset which you will track through the course of this project.
Run the code block below to observe a statistical description of the dataset. Note that the dataset is composed of six important product categories: 'Fresh', 'Milk', 'Grocery', 'Frozen', 'Detergents_Paper', and 'Delicatessen'. Consider what each category represents in terms of products you could purchase.
In [3]:
# Display a description of the dataset
display(data.describe())
In [4]:
print data.loc[data['Milk']>20000]
To get a better understanding of the customers and how their data will transform through the analysis, it would be best to select a few sample data points and explore them in more detail. In the code block below, add three indices of your choice to the indices list which will represent the customers to track. It is suggested to try different sets of samples until you obtain customers that vary significantly from one another.
In [5]:
# TODO: Select three indices of your choice you wish to sample from the dataset
indices = [100,183,309]
# Create a DataFrame of the chosen samples
samples = pd.DataFrame(data.loc[indices], columns = data.keys()).reset_index(drop = True)
print "Chosen samples of wholesale customers dataset:"
display(samples)
In [6]:
for column in data.columns:
print column + " " + str(data[column].mean())
In [26]:
# Code from reviewer suggestion
samples_bar = samples.append(data.describe().loc['mean'])
samples_bar.index = indices + ['mean']
_ = samples_bar.plot(kind='bar', figsize=(14,6))
Consider the total purchase cost of each product category and the statistical description of the dataset above for your sample customers.
What kind of establishment (customer) could each of the three samples you've chosen represent?
Hint: Examples of establishments include places like markets, cafes, and retailers, among many others. Avoid using names for establishments, such as saying "McDonalds" when describing a sample customer as a restaurant.
Answer:
Sample 1 has above average spending on all the features except for Detergents, but spends the most on Delicatessen, this could be a large supermarket with a deli department. Or it could also be a big deli restaurant given the low spending on detergents. Clearly they are expecting people to buy food related products compared to cleaning supplies.
Sample 2 has much below average spending on fresh product,relatively high spending on Milk and grocery, unusually low spending on Frozen and Delicatessen and pretty high spending on Detergents. Given high spenidng on Milk, grocery and Detergents I think it's a small supermarket with a bakery department.
One interesting thought to consider is if one (or more) of the six product categories is actually relevant for understanding customer purchasing. That is to say, is it possible to determine whether customers purchasing some amount of one category of products will necessarily purchase some proportional amount of another category of products? We can make this determination quite easily by training a supervised regression learner on a subset of the data with one feature removed, and then score how well that model can predict the removed feature.
In the code block below, you will need to implement the following:
new_data a copy of the data by removing a feature of your choice using the DataFrame.drop function.sklearn.cross_validation.train_test_split to split the dataset into training and testing sets.test_size of 0.25 and set a random_state.random_state, and fit the learner to the training data.score function.
In [7]:
# Trying with dropping all the features one by one and checking the R^S score.
from sklearn.cross_validation import train_test_split
from sklearn.tree import DecisionTreeRegressor
columns = data.columns
for column in columns:
# TODO: Make a copy of the DataFrame, using the 'drop' function to drop the given feature
new_data = data.drop(column,axis = 1)
target_label = data[column]
# TODO: Split the data into training and testing sets using the given feature as the target
X_train, X_test, y_train, y_test = train_test_split(new_data,target_label,test_size = 0.25, random_state = 0)
# TODO: Create a decision tree regressor and fit it to the training set
regressor = DecisionTreeRegressor(random_state = 0)
regressor.fit(X_train,y_train)
# TODO: Report the score of the prediction using the testing set
score = regressor.score(X_test,y_test)
print "When Removed Feature is " + str(column) + " R^2 score is " + str(round(score,3))
Based on the R^2 scores, I choose Delicatessen as the dropped feature.
Which feature did you attempt to predict? What was the reported prediction score? Is this feature is necessary for identifying customers' spending habits?
Hint: The coefficient of determination, R^2, is scored between 0 and 1, with 1 being a perfect fit. A negative R^2 implies the model fails to fit the data.
Answer:
Feature chosen to drop was Delicatessen and it appears the Decision Tree Regressor failed to predict from the purchase of other products how much a particular customer will spend on Deli products. If we drop this feature, we will lose information given spending on other features did not have much correlation with this feature. The reported prediction score was R^2 = -11.664 which means the model absolutely failed to fit the data.We can say that without this feature, we will never know from the other features how much a customer would have spent on Deli products, given lack of correlation between this feature and the other one.
To get a better understanding of the dataset, we can construct a scatter matrix of each of the six product features present in the data. If you found that the feature you attempted to predict above is relevant for identifying a specific customer, then the scatter matrix below may not show any correlation between that feature and the others. Conversely, if you believe that feature is not relevant for identifying a specific customer, the scatter matrix might show a correlation between that feature and another feature in the data. Run the code block below to produce a scatter matrix.
In [8]:
# Produce a scatter matrix for each pair of features in the data
pd.scatter_matrix(data, alpha = 0.3, figsize = (14,8), diagonal = 'kde');
Are there any pairs of features which exhibit some degree of correlation? Does this confirm or deny your suspicions about the relevance of the feature you attempted to predict? How is the data for those features distributed?
Hint: Is the data normally distributed? Where do most of the data points lie?
Answer:
From the visualization the observed pairs are : 1. Grocery and Detergent paper, 2. Grocery and Milk, 3. Milk and Detergent Paper.
These pairings are not really relevant for my dropped feature that I attempted to predict given there's not much correlation between the dropped feature Delicatessen and the other features. Most consumers spend relatively low amount on Delicatessen with an average of 1524.87 m.u (monetary unit) with one outlier who purchase a very high amount of deli products(above 40000) and a few other one's above 10000. The visualization does confirm my R^2 that Deli products are quite uncorrelated with other products as the purchase of deli does not really increase with the purchase of other features and neither it decreases with the purchase of other features. It's just relatively low, presumably because most retailers don't have a deli department, or even if they have one, it's a small one. The data does not seem to be normally distributed, rather it's mostly left skewed and the median falls below the mean.
In this section, you will preprocess the data to create a better representation of customers by performing a scaling on the data and detecting (and optionally removing) outliers. Preprocessing data is often times a critical step in assuring that results you obtain from your analysis are significant and meaningful.
If data is not normally distributed, especially if the mean and median vary significantly (indicating a large skew), it is most often appropriate to apply a non-linear scaling — particularly for financial data. One way to achieve this scaling is by using a Box-Cox test, which calculates the best power transformation of the data that reduces skewness. A simpler approach which can work in most cases would be applying the natural logarithm.
In the code block below, you will need to implement the following:
log_data after applying a logarithm scaling. Use the np.log function for this.log_samples after applying a logrithm scaling. Again, use np.log.
In [9]:
# TODO: Scale the data using the natural logarithm
log_data = np.log(data)
# TODO: Scale the sample data using the natural logarithm
log_samples = np.log(samples)
# Produce a scatter matrix for each pair of newly-transformed features
pd.scatter_matrix(log_data, alpha = 0.3, figsize = (14,8), diagonal = 'kde');
After applying a natural logarithm scaling to the data, the distribution of each feature should appear much more normal. For any pairs of features you may have identified earlier as being correlated, observe here whether that correlation is still present (and whether it is now stronger or weaker than before).
Run the code below to see how the sample data has changed after having the natural logarithm applied to it.
In [10]:
# Display the log-transformed sample data
display(log_samples)
Detecting outliers in the data is extremely important in the data preprocessing step of any analysis. The presence of outliers can often skew results which take into consideration these data points. There are many "rules of thumb" for what constitutes an outlier in a dataset. Here, we will use Tukey's Method for identfying outliers: An outlier step is calculated as 1.5 times the interquartile range (IQR). A data point with a feature that is beyond an outlier step outside of the IQR for that feature is considered abnormal.
In the code block below, you will need to implement the following:
Q1. Use np.percentile for this.Q3. Again, use np.percentile.step.outliers list.NOTE: If you choose to remove any outliers, ensure that the sample data does not contain any of these points!
Once you have performed this implementation, the dataset will be stored in the variable good_data.
In [11]:
# For Counting how many times each indices appear as outliers
frequent_outlier_indices = { }
# For each feature find the data points with extreme high or low values
for feature in log_data.keys():
# TODO: Calculate Q1 (25th percentile of the data) for the given feature
Q1 = np.percentile(log_data[feature],25)
# TODO: Calculate Q3 (75th percentile of the data) for the given feature
Q3 = np.percentile(log_data[feature],75)
# TODO: Use the interquartile range to calculate an outlier step (1.5 times the interquartile range)
step = 1.5*(Q3-Q1)
# Display the outliers
print "Data points considered outliers for the feature '{}':".format(feature)
outlier_dataframe = log_data[~((log_data[feature] >= Q1 - step) & (log_data[feature] <= Q3 + step))]
display(outlier_dataframe)
outlier_index_list = list(outlier_dataframe.index.values)
for index in outlier_index_list:
if index in frequent_outlier_indices:
frequent_outlier_indices[index]+=1
else:
frequent_outlier_indices[index]=1
#Only keep indices which occur more than once
frequent_outlier_indices = {index:value for index,value in frequent_outlier_indices.iteritems() if value>1}
print frequent_outlier_indices
In [12]:
# OPTIONAL: Select the indices for data points you wish to remove
outliers= [key for key in frequent_outlier_indices]
display(data.ix[outliers])
# Remove the outliers, if any were specified
good_data = log_data.drop(log_data.index[outliers]).reset_index(drop = True)
In [13]:
print data.shape
print good_data.shape
Answer:
Yes, the indices 128,65,75,66,154 was considered for more than one feature. The data frame for these indices can be seen above.
outliers list to be removed, explain why.I decided to remove all the data points from this data set because these outliers may skew the direction of the ordered principal components which should be pointed at the direction of maximum variance of the data.PCA tries to minimize the information loss(measured by the distance from the points to their new projected spots in the new feature), however these outliers add noise to the dataset and skew the variance so the PCA wouldn't be able to output the principal components which would point to the direction of maximum variance. If we try to work on clustering on this dataset these outliers might end up in their own clusters of one-two points which wiil not be informative as their dissimilarity(measured by average distance of the means or some other distance based metric) will be too big compared to other data points. To be clear, it's possible that removing points that has only one feature as outlier may skew the variance, but may be it'd be a good idea to iterate the similar experiments with only a few of the points removed and checking how that works in future. We can iteratively remove outliers one by one and check our results in a future experiment.
In this section you will use principal component analysis (PCA) to draw conclusions about the underlying structure of the wholesale customer data. Since using PCA on a dataset calculates the dimensions which best maximize variance, we will find which compound combinations of features best describe customers.
Now that the data has been scaled to a more normal distribution and has had any necessary outliers removed, we can now apply PCA to the good_data to discover which dimensions about the data best maximize the variance of features involved. In addition to finding these dimensions, PCA will also report the explained variance ratio of each dimension — how much variance within the data is explained by that dimension alone. Note that a component (dimension) from PCA can be considered a new "feature" of the space, however it is a composition of the original features present in the data.
In the code block below, you will need to implement the following:
sklearn.decomposition.PCA and assign the results of fitting PCA in six dimensions with good_data to pca.log_samples using pca.transform, and assign the results to pca_samples.
In [14]:
from sklearn.decomposition import PCA
# TODO: Apply PCA by fitting the good data with the same number of dimensions as features
n_features = len(good_data.columns)
pca = PCA(n_components = n_features)
pca.fit(good_data)
# TODO: Transform the sample log-data using the PCA fit above
pca_samples = pca.transform(log_samples)
# Generate PCA results plot
pca_results = rs.pca_results(good_data, pca)
How much variance in the data is explained in total by the first and second principal component? What about the first four principal components? Using the visualization provided above, discuss what the first four dimensions best represent in terms of customer spending.
Hint: A positive increase in a specific dimension corresponds with an increase of the positive-weighted features and a decrease of the negative-weighted features. The rate of increase or decrease is based on the indivdual feature weights.
How much variance in the data is explained by the first and second principal component?
0.4430 + 0.2638 = 0.7068 or 70.68% variance in data is explained by the first and second principal component.
What about the first four principal components?
(0.4430 + 0.2638 + 0.1231 + 0.1012) = 0.9319 or 93.11% of the total variance in data is explained by the first four principal components.
Using the visualization provided above, discuss what the first four dimensions best represent in terms of customer spending.
The first principal component shows strong positive correlation with detergents, grocery and milk, small positive correlation with deli and low negative correlation with fresh and frozen. This shows some consumers who spend a lot on detergents, also tend to spend a lot on grocery and milk, and slightly on deli , however they spend less on fresh and frozen food. This may represent retailers who don't focus much on fresh or frozen food distribution, but generally distributes grocery and milk.It can also be hotels which mostly buy groceries and milk for producing meals, which may explain why there's a lack of frozen food buying.
The second principal component shows strong positive correlation with fresh, frozen and deli and slightly positive correlation with the other three. It apprears that this group is absolutely different from the first component as in the first component fresh and frozen spending was relatively low. It shows that some consumers tend to spend a lot on fresh, frozen and deli products, however they might also buy some detergent, grocery and milk at the same time. This is probably a retailer focusing on selling fresh, frozen and deli products. Deli products is somewhat unusual still to sale and a retailer who'd focus on selling deli, may pick up the idea that many consumers like fresh, organic products while some consumers buy frozen food for convinience and buy them for reselling.
The third principal component shows strong positive correlation with deli products and moderate positve correlation with frozen products, but it also shows strong negative correlation with fresh and detergents. It seems that some consumers who spend a lot on deli products, may spend a lot on frozen products too, but they will spend a lot less on fresh and detergents. This can represent deli restaurants, because retailers and hotels might buy more detergents which would end up showing a positive correlation with that feature.
The fourth principal component shows a very strong positive correlation with frozen feature and a small positive correlation with detergents, but it has strong negative correlation with deli products and small negative correlation with fresh feature. So, it seems some consumers who spend a lot on frozen product and a little bit on detergents spends a lot less on the deli products and slightly less on fresh products. It may explain the difference between supermarkets/retailers with a deli department who spend a lot on frozen foods and other items vs the one's who don't have a deli department. It's probably a retailer who sales a lot of frozen products, some grocery and other products but does not sale deli or fresh products at all.
Resource used : https://onlinecourses.science.psu.edu/stat505/node/54
Run the code below to see how the log-transformed sample data has changed after having a PCA transformation applied to it in six dimensions. Observe the numerical value for the first four dimensions of the sample points. Consider if this is consistent with your initial interpretation of the sample points.
In [15]:
# Display sample log-data after having a PCA transformation applied
display(pd.DataFrame(np.round(pca_samples, 4), columns = pca_results.index.values))
When using principal component analysis, one of the main goals is to reduce the dimensionality of the data — in effect, reducing the complexity of the problem. Dimensionality reduction comes at a cost: Fewer dimensions used implies less of the total variance in the data is being explained. Because of this, the cumulative explained variance ratio is extremely important for knowing how many dimensions are necessary for the problem. Additionally, if a signifiant amount of variance is explained by only two or three dimensions, the reduced data can be visualized afterwards.
In the code block below, you will need to implement the following:
good_data to pca.good_data using pca.transform, and assign the reuslts to reduced_data.log_samples using pca.transform, and assign the results to pca_samples.
In [16]:
# TODO: Apply PCA by fitting the good data with only two dimensions
pca = PCA(n_components = 2)
pca.fit(good_data)
# TODO: Transform the good data using the PCA fit above
reduced_data = pca.transform(good_data)
# TODO: Transform the sample log-data using the PCA fit above
pca_samples = pca.transform(log_samples)
# Create a DataFrame for the reduced data
reduced_data = pd.DataFrame(reduced_data, columns = ['Dimension 1', 'Dimension 2'])
In [17]:
# Display sample log-data after applying PCA transformation in two dimensions
display(pd.DataFrame(np.round(pca_samples, 4), columns = ['Dimension 1', 'Dimension 2']))
In this section, you will choose to use either a K-Means clustering algorithm or a Gaussian Mixture Model clustering algorithm to identify the various customer segments hidden in the data. You will then recover specific data points from the clusters to understand their significance by transforming them back into their original dimension and scale.
Answer:
K means clustering is a hard clustering method(each instance gets assigned to only one of the clusters instead of generating probabilities for each cluster). It's easy to implement, simple to understand and depending on the domain by selecting good distance metric(for measuring similarity) practioners can try many different variants of k-means to get good results.
Gaussian Mixture Models is a soft clustering technique(for each instance we generate the probability of it belonging to the different classes). We assume the data points are sourced from a mixture of finite number of gaussian distributions with unknown parameters and for each instance we figure out the probability of it's being from different gaussians after estimating those parameters. It assumes a point can be shared by two clusters, which is often more realistic as an assumption as many instances will have shared characteristics. Also by getting the probabilities it's easier to see the boundary cases, instances which could have belonged to either clusters which is more useful for industry decision making.
I think I'll use the Gaussian mixture model because if the distributor gets the boundary cases as well as the probability of each customer belonging to different clusters, it might be more helpful to segment them, they may decide to take the boundary cases to the closest cluster, but at least it'd be an informed decision. K means is also a special case of Gaussian Mixture models so I'm assuming if the datapoint falls clearly into a cluster and there's no boundary cases GMM will pick it up, while Kmeans will hard assign the instance to some cluster, so it's a good idea to check for edge cases given the business revenue will be impacted by the segmentation methods in future.
Resources used : http://scikit-learn.org/stable/modules/mixture.html#gmm-classifier , https://www.quora.com/What-is-the-difference-between-K-means-and-the-mixture-model-of-Gaussian , https://www.quora.com/What-is-an-intuitive-explanation-of-Gaussian-mixture-models
Depending on the problem, the number of clusters that you expect to be in the data may already be known. When the number of clusters is not known a priori, there is no guarantee that a given number of clusters best segments the data, since it is unclear what structure exists in the data — if any. However, we can quantify the "goodness" of a clustering by calculating each data point's silhouette coefficient. The silhouette coefficient for a data point measures how similar it is to its assigned cluster from -1 (dissimilar) to 1 (similar). Calculating the mean silhouette coefficient provides for a simple scoring method of a given clustering.
In the code block below, you will need to implement the following:
reduced_data and assign it to clusterer.reduced_data using clusterer.predict and assign them to preds.centers.pca_samples and assign them sample_preds.reduced_data against preds.score and print the result.
In [18]:
from sklearn.mixture import GMM
from sklearn.metrics import silhouette_score
scores = []
for n in xrange(2,11):
# TODO: Apply your clustering algorithm of choice to the reduced data
clusterer = GMM(n_components = n,random_state = 0)
clusterer.fit(reduced_data)
# TODO: Predict the cluster for each data point
preds = clusterer.predict(reduced_data)
# TODO: Find the cluster centers
centers = clusterer.means_
# TODO: Predict the cluster for each transformed sample data point
sample_preds = clusterer.predict(pca_samples)
# TODO: Calculate the mean silhouette coefficient for the number of clusters chosen
score = silhouette_score(reduced_data,preds)
scores.append(score)
print "For cluster number "+ str(n) + " the score is " + str(score)
In [19]:
import matplotlib.pyplot as plt
components = list(xrange(2,11))
plt.plot(components,scores)
plt.xlabel("Number of clusters")
plt.ylabel("Silhoutte score")
Out[19]:
Answer:
Cluster number 2 has the best silhoutte score so I'm choosing it for the final model.
In [20]:
# TODO: Apply your clustering algorithm of choice to the reduced data
clusterer = GMM(n_components = 2,random_state = 0)
clusterer.fit(reduced_data)
# TODO: Predict the cluster for each data point
preds = clusterer.predict(reduced_data)
# TODO: Find the cluster centers
centers = clusterer.means_
# TODO: Predict the cluster for each transformed sample data point
sample_preds = clusterer.predict(pca_samples)
# TODO: Calculate the mean silhouette coefficient for the number of clusters chosen
score = silhouette_score(reduced_data,preds)
Once you've chosen the optimal number of clusters for your clustering algorithm using the scoring metric above, you can now visualize the results by executing the code block below. Note that, for experimentation purposes, you are welcome to adjust the number of clusters for your clustering algorithm to see various visualizations. The final visualization provided should, however, correspond with the optimal number of clusters.
In [21]:
# Display the results of the clustering from implementation
rs.cluster_results(reduced_data, preds, centers, pca_samples)
Each cluster present in the visualization above has a central point. These centers (or means) are not specifically data points from the data, but rather the averages of all the data points predicted in the respective clusters. For the problem of creating customer segments, a cluster's center point corresponds to the average customer of that segment. Since the data is currently reduced in dimension and scaled by a logarithm, we can recover the representative customer spending from these data points by applying the inverse transformations.
In the code block below, you will need to implement the following:
centers using pca.inverse_transform and assign the new centers to log_centers.np.log to log_centers using np.exp and assign the true centers to true_centers.
In [22]:
# TODO: Inverse transform the centers
log_centers = pca.inverse_transform(centers)
# TODO: Exponentiate the centers
true_centers = np.exp(log_centers)
# Display the true centers
segments = ['Segment {}'.format(i) for i in range(0,len(centers))]
true_centers = pd.DataFrame(np.round(true_centers), columns = data.keys())
true_centers.index = segments
display(true_centers)
In [23]:
# Display a description of the dataset
display(data.describe())
Consider the total purchase cost of each product category for the representative data points above, and reference the statistical description of the dataset at the beginning of this project. What set of establishments could each of the customer segments represent?
Hint: A customer who is assigned to 'Cluster X' should best identify with the establishments represented by the feature set of 'Segment X'.
In [27]:
true_centers = true_centers.append(data.describe().loc['mean'])
_ = true_centers.plot(kind='bar', figsize=(15,6))
Answer:
I think the segments basically represent two sorts of customers, customers who spend below average on every feature(segment 0) and the customers who spend near average or above average on the features available(segment 1). I believe segment 0 represents smaller consumers such as restaurants or cafe's, while the segment 1 generally represents big establishments such as supermarkets or retailers. The retailers may differ from each other on the basis of spending differently on frozen or deli products, but they spend higher than average on the fresh, grocery and milk compared to the segment 0 which are smaller establishments.
As we see from the visualization, average customers in segment 0 has below the mean spending on all features while the average customer spending in segment 1 looks much similar to the mean spending. Given the data is left skewed, it is expected there is a subgroup of consumers with high spending on all features.
In [24]:
# Display the predictions
for i, pred in enumerate(sample_preds):
print "Sample point", i, "predicted to be in Cluster", pred
Answer:
This is my first take on the samples (from above)
Sample 0 has average spending on Fresh food, slightly above average on milk and grocery, similar to average spending on Frozen but unusually high spending on detergent paper, it could be a hotel with many rooms which require a lot of cleaning supplies but has an in-house restaurant , or it could be a retailer which supplies cleaning products to other smaller establishments.
Sample 1 has above average spending on all the features except for Detergents, but spends the most on Delicatessen, this could be a large supermarket with a deli department. Or it could also be a big deli restaurant given the low spending on detergents. Clearly they are expecting people to buy food related products compared to cleaning supplies.
Sample 2 has much below average spending on fresh product,relatively high spending on Milk and grocery, unusually low spending on Frozen and Delicatessen and pretty high spending on Detergents. Given high spenidng on Milk, grocery and Detergents I think it's a small supermarket with a bakery department.
I assumed my samples differ significantly from each other because they spent relatively very high or low on one feature, such as detergent paper or deli products or milk. However the clustering predicts all of them are from segment one i.e all of them are retailers. I think this is actually consistent with my hypotheses that all of them are some variations of hotels or supermarket. And it appears that the clustering is just segmenting the smaller establishments from the bigger ones which can actually be useful in practice.
It's true that I assumed sample 0 is different from sample 1,2 because it spent a lot on detergent products, I also assumed sample 1 is different from sample 0 and 2 because it spent a lot on deli products. Sample 2 had a lot below average spending on the fresh product, but it's also possible it's just a recording error. Even with relative differences, on the whole they were consumers who spent quite high on all the features compared to smaller establishments.
Perhaps with different samples we'd be able to take a look at the custoemrs from segment 0, but overall I think the results are consistent. I also think choosing samples who spend a lot on some dimension was a good choice still because now I know after the clustering that even if some consumers differ a lot from each other in superficial aspects, such as high spending on one dimension, they still differ from the smaller establishments when it comes to average spending on all the features.
Companies often run A/B tests when making small changes to their products or services. If the wholesale distributor wanted to change its delivery service from 5 days a week to 3 days a week, how would you use the structure of the data to help them decide on a group of customers to test?
Hint: Would such a change in the delivery service affect all customers equally? How could the distributor identify who it affects the most?
Answer:
For each segment we can randomly divide the customers into two groups, control(group A) and variation group(group B). Group A will get the usual 5 days a week service while the group B will get the 3 days a week service. My assumption is that the consumers from segment 0(the smaller establishments) may be comfortable with the 3 days a week service, while the customers from segment 1 (big retailers/hotels) will probably need 5 days a week service for frequent refills, but only after comparing the feedback from the control and variation groups of each segment we can decide. Given customers in a segment are similar to each other, the distributor should compare the feedbacks from the control vs variation group between a segment, and how each segment differs from each other as a whole too given larger establishments may not be comfortable with the 3 days service and might be prone to switching. For getting feedback on the change, we can use survey questions or track the purchasing behavior of the control vs the variation groups.
Assume the wholesale distributor wanted to predict a new feature for each customer based on the purchasing information available. How could the wholesale distributor use the structure of the clustering data you've found to assist a supervised learning analysis?
Hint: What other input feature could the supervised learner use besides the six product features to help make a prediction?
Answer:
Besides the six product features, perhaps whether a customer is from segment 1 or 0, e.g (smaller or bigger establishment/below average vs equal/higher than average spending) can be used as a feature and in fact it may be a very important categorical variable.
At the beginning of this project, it was discussed that the 'Channel' and 'Region' features would be excluded from the dataset so that the customer product categories were emphasized in the analysis. By reintroducing the 'Channel' feature to the dataset, an interesting structure emerges when considering the same PCA dimensionality reduction applied earlier on to the original dataset.
Run the code block below to see how each data point is labeled either 'HoReCa' (Hotel/Restaurant/Cafe) or 'Retail' the reduced space. In addition, you will find the sample points are circled in the plot, which will identify their labeling.
In [25]:
# Display the clustering results based on 'Channel' data
rs.channel_results(reduced_data, outliers, pca_samples)
How well does the clustering algorithm and number of clusters you've chosen compare to this underlying distribution of Hotel/Restaurant/Cafe customers to Retailer customers? Are there customer segments that would be classified as purely 'Retailers' or 'Hotels/Restaurants/Cafes' by this distribution? Would you consider these classifications as consistent with your previous definition of the customer segments?
Answer:
The clustering algorithm chose had 2 clusters. In the underlying distribution there's also two clusters hotels/cafes/restaurants vs retailers. My clustering model segmented based on smaller vs larger spending compared to average, and in the underlying distribution the cafes or restaurants will spend much less compared to a retailer.
I think aside from the boundary cases between cluster 0 and 1, the red points in my original clustering are clearly hotels/restaurants/cafe's and the purple points in cluster 1 are mostly retailers.
Interestingly enough, I chose the sample 1 as a retailer and it seems my clustering predicted it in cluster 1(retailers) while it's actually a hotel in the original distribution, but even if it's a hotel it's clearly near the boundary of the cluster 1, which makes it much similar to retailers insted of the hotels. I'm wondering if there's some recording error in the original data too, given some green points/points identified retailers are clearly similar to hotels. But it's possible that the retailers for the small regions, depending on the size also spends simlar to cafes/restaurants and some big hotels or restaurant may spend similar to the average retailers.
Note: Once you have completed all of the code implementations and successfully answered each question above, you may finalize your work by exporting the iPython Notebook as an HTML document. You can do this by using the menu above and navigating to
File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In [ ]: