In this lesson, we will examine the KMeans algorithum using the Scikit-Learning Library. Clustering algorithums are unsupervised algorithums, where the training data is not labeled. The algorithums cluster or group data sets based on common characteristics. We are going to look at sales figures,gender, age for our customers. Our company is an online clothing store for men. We are going to look at the data we have collected on age, gender and sales for each of our 200 customers. Your first assignment is to come up with a spending score based on sales for each customer. I have created a Python program that evaluates indicidual sales amounts for each customer and assigns a spending score to it. The sales represent sales for the month. Sales range for as little as $5.00 to a maximum of $500.

Day 1 & 2: Assigning spending scores to the spreadsheet for each of our 200 customers.

Here are the sales for each customer.

Customer Id	Monthly Sales
1	196.35
2	406.95
3	29.99
4	387.95
5	204.45
6	380.95
7	29.99
8	373.25
9	5.95
10	362.95
11	65.99
12	495.95
13	73.99
14	387.95
15	64.35
16	395.95
17	175.95
18	333.95
19	144.95
20	492.25
21	175.95
22	365.95
23	22.50
24	365.95
25	65.99
26	411.95
27	411.95
28	161.95
29	308.95
30	150.95
31	438.95
32	15.37
33	365.95
34	15.37
35	436.25
36	65.99
37	406.95
38	81.45
39	365.95
40	125.95
41	379.95
42	175.95
43	463.25
44	183.55
45	308.95
46	136.95
47	275.95
48	235.95
49	212.45
50	212.45
51	262.95
52	304.25
53	273.95
54	304.25
55	227.45
56	206.45
57	253.95
58	234.95
59	255.95
60	234.95
61	280.45
62	275.95
63	262.95
64	295.25
65	255.95
66	295.25
67	253.93
68	240.95
69	295.25
70	235.95
71	275.95
72	212.45
73	248.95
74	280.45
75	235.95
76	273.95
77	268.95
78	240.95
79	262.95
80	212.45
81	255.95
82	275.95
83	206.45
84	224.45
85	285.45
86	234.95
87	291.25
88	275.95
89	304.25
90	234.95
91	275.95
92	218.45
93	248.95
94	204.45
95	212.45
96	262.95
97	235.95
98	253.95
99	212.45
100	248.95
101	206.45
102	240.95
103	295.25
104	275.95
105	280.45
106	212.45
107	253.95
108	234.95
109	218.45
110	240.95
111	262.95
112	273.95
113	212.45
114	234.95
115	240.95
116	253.95
117	218.45
118	295.25
119	218.45
120	285.45
121	280.45
122	204.45
123	291.25
124	455.95
125	144.95
126	387.95
127	175.95
128	475.25
129	53.95
130	379.95
131	43.95
132	379.95
133	172.95
134	356.95
135	22.50
136	442.95
137	32.67
138	365.95
139	48.95
140	362.95
141	22.50
142	465.25
143	204.45
144	438.95
145	59.35
146	486.25
147	183.55
148	373.95
149	108.45
150	450.95
151	81.45
152	442.95
153	96.0
154	380.95
155	78.95
156	447.95
157	2.50
158	390.95
159	2.50
160	365.95
161	175.95
162	417.95
163	22.50
164	465.25
165	125.95
166	379.95
167	96.00
168	475.25
169	134.95
170	317.95
171	64.35
172	379.95
173	48.95
174	463.25
175	64.35
176	433.95
177	73.99
178	347.95
179	65.99
180	450.95
181	161.95
182	433.95
183	73.99
184	442.95
185	197.75
186	486.25
187	118.45
188	340.95
189	81.45
190	427.95
191	110.45
192	347.95
193	39.50
194	455.95
195	78.95
196	395.95
197	136.95
198	373.95
199	87.45
200	417.95

Look at the chart in the next column to find their spending scores.

Spending Scores

if sales <=10:
    spendingScore = 3   
   
if sales>10 and sales<=20:
    spendingScore = 4

if sales>20 and sales<=25:
    spendingScore = 5

if sales>25 and sales<=30:
    spendingScore = 6
        
if sales>30 and sales<=35:
    spendingScore = 7

if sales>35 and sales<=40:
    spendingScore = 8

if sales>40 and sales<=45:
    spendingScore = 9

if sales>45 and sales<=50:
    spendingScore = 10

if sales>50 and sales<=55:
    spendingScore = 11
  
if sales>55 and sales<=60:
    spendingScore = 12

if sales>60 and sales<=65:
    spendingScore = 13

if sales>65 and sales<=70:
    spendingScore = 14

if sales>70 and sales<=75:
    spendingScore = 15

if sales>75 and sales<=80:
    spendingScore = 16

if sales>80 and sales<=85:
    spendingScore = 17

if sales>85 and sales<=90:
    spendingScore = 18

if sales>90 and sales<=95:
    spendingScore = 19
      
if sales>95 and sales<=100:
    spendingScore = 20
         
if sales>100 and sales<=105:
    spendingScore = 21
              
if sales>105 and sales<=110:
    spendingScore = 22
    
if sales>110 and sales<=115:
    spendingScore = 23
     
if sales>115 and sales<=120:
    spendingScore = 24
      
if sales>120 and sales<=125:
    spendingScore = 25

if sales>125 and sales<=130:
    spendingScore = 26

if sales>130 and sales<=135:
    spendingScore = 27

if sales>135 and sales<=140:
    spendingScore = 28

if sales>140 and sales<=145:
    spendingScore = 29

if sales>145 and sales<=150:
    spendingScore = 30

if sales>150 and sales<=155:
    spendingScore = 31

if sales>160 and sales<=165:
    spendingScore = 32

if sales>165 and sales<=170:
    spendingScore = 33

if sales>170 and sales<=175:
    spendingScore = 34

if sales>175 and sales<=180:
    spendingScore = 35

if sales>180 and sales<=185:
    spendingScore = 36

if sales>185 and sales<=190:
    spendingScore = 37

if sales>190 and sales<=195:
    spendingScore = 38

if sales>195 and sales<=200:
    spendingScore = 39

if sales>200 and sales<=205:
    spendingScore = 40

if sales>205 and sales<=210:
    spendingScore = 41

if sales>210 and sales<=215:
    spendingScore = 42

if sales>215 and sales<=220:
    spendingScore = 43

if sales>220 and sales<=225:
    spendingScore = 44

if sales>225 and sales<=230:
    spendingScore = 45

if sales>230 and sales<=235:
    spendingScore = 46

if sales>235 and sales<=240:
    spendingScore = 47
    
if sales>240 and sales<=245:
    spendingScore = 48
    
if sales>245 and sales<=250:
    spendingScore = 49
    
if sales>250 and sales<=255:
    spendingScore = 50
    
if sales>255 and sales<=260:
    spendingScore = 51
    
if sales>260 and sales<=265:
    spendingScore = 52
    
if sales>265 and sales<=270:
    spendingScore = 53
    
if sales>270 and sales<=275:
    spendingScore = 54
    
if sales>275 and sales<=280:
    spendingScore = 55
    
if sales>280 and sales<=285:
    spendingScore = 56
    
if sales>285 and sales<=290:
    spendingScore = 57
    
if sales>290 and sales<=295:
    spendingScore = 58
    
if sales>295 and sales<=300:
    spendingScore = 59
    
if sales>300 and sales<=305:
    spendingScore = 60
    
if sales>305 and sales<=310:
    spendingScore = 61
    
if sales>310 and sales<=315:
    spendingScore = 62
    
if sales>315 and sales<=320:
    spendingScore = 63
    
if sales>325 and sales<=330:
    spendingScore = 65
    if sales>330 and sales<=335:
    spendingScore = 66
    if sales>335 and sales<=340:
    
if sales>340 and sales<=345:
    spendingScore = 68
    
if sales>345 and sales<=350:
    spendingScore = 69
    
if sales>350 and sales<=355:
    spendingScore = 70
    
if sales>355 and sales<=360:
    spendingScore = 71
    
if sales>360 and sales<=365:
    spendingScore = 72
    
if sales>365 and sales<=370:
    spendingScore = 73
    
if sales>370 and sales<=375:
    spendingScore = 74
    
if sales>375 and sales<=380:
    spendingScore = 75
    
if sales>380 and sales<=385:
    spendingScore = 76
    
if sales>385 and sales<=390:
    spendingScore = 77
    
if sales>390 and sales<=395:
    spendingScore = 78
    
if sales>395 and sales<=400:
    spendingScore = 79
    
if sales>400 and sales<=405:
    spendingScore = 80
    
if sales>405 and sales<=410:
    spendingScore = 81
    
if sales>410 and sales<=415:
    spendingScore = 82
    
if sales>415 and sales<=420:
    spendingScore = 83
    
if sales>420 and sales<=425:
    spendingScore = 84
    
if sales>425 and sales<=430:
    spendingScore = 85
    
if sales>430 and sales<=435:
    spendingScore = 86
    
if sales>435 and sales<=440:
    spendingScore = 87
    
if sales>440 and sales<=445:
    spendingScore = 88
    
if sales>445 and sales<=450:
    spendingScore = 89
    
if sales>450 and sales<=455:
    spendingScore = 90
    
if sales>455 and sales<=460:
    spendingScore = 91
    
if sales>460 and sales<=465:
    spendingScore = 92
    
if sales>465 and sales<=470:
    spendingScore = 93
    
if sales>470 and sales<=475:
    spendingScore = 94
    
if sales>475 and sales<=480:
    spendingScore = 95
    
if sales>480 and sales<=485:
    spendingScore = 96
    
if sales>485 and sales<=490:
    spendingScore = 97
    
if sales>490 and sales<=495:
    spendingScore = 98
    
if sales>495 and sales<=500:
    spendingScore = 99
    
if sales>500:
    spendingScore = 100

Open the spreadsheet file
customers
Look up the scores for each customer based on their sales.
Put spending scores in for each customer 1-200 using column E.
Save your file as an csv file in your working folder and call it "customers.csv"

Day 3: Python coding:Getting the libraries and loading the file.

Open a new Python project, click copy text button and paste the contents into the first frame.

Run the code in the first frame.

This code imports libraries from Python, reads in the spreadsheet file and prints the entire dataset.

You will have to adjust the location of the file in the code to your folder where you saved the cvs spreadsheet file.

When you run the code, you should see the shape of the data set which contains 200 rows and 5 columns.

The entire datset should also shows: customer id, gender,age, annual income and spending score.

Create a new frame by clicking on the + icon and add this line: sns.histplot(dataset['Annual Income (k$)'], kde=False, bins=50)

Your output should be a graph of the annual income. As you can see most of our customers have and annual income between $60,000 and $90,000 per year.

Now let create a histogram showing spending scores. Create a new frame and key in this code: sns.histplot(dataset['Spending Score (1-100)'],kde=False, bins=50 ,color="red")

The histogram shows that the spending scores are between 40 and 60.

Now let's see if there is a linear relationship between annual income and spending. Key in these lines of code to generate a scatter plot.

#plotting regression plot for annual income against spending score
sns.regplot(x ="Annual Income (k$)", y = "Spending Score (1-100)",data = dataset)

From the straight line, you can determine that there is no linear relationship between annual income and sending.

Now let's look to see if there is a linear relationshp between age and spending score. Create a new frame and add these lines.

#plotting regression plot for age aganst spending score
sns.regplot(x="Age", y="Spending Score (1-100)",data = dataset)

Our scatter plot shows an inverse linear relationship between age and spending score. Young people have higher spending scores compared to older folks.

Day 4:Customer Segmentation

Our goal here is to target a customer with high income and a high spending score. We want to filter the spending score and annual income and eliminate the other columns in the dataset. Here is the code to accomplish this.

Create a new frame and click the copy text button and paste the contents into the frame.

Save your program the run the code in frame.

Here is the output that you should get.

To implement K-Means clustering, you can use the k-Means class from the sklearn.cluster module of the Sklearn library. You have to pass the number of clusters as an attribute to the k-Means class constructor. To train the K-Means model, simply pass the dataset to the fit() method of the K-Means class.¹ #performing kmeans clustering using kmeans class # you guess that there are 4 clusters km_model = KMeans(n_clusters=4) km_model.fit(dataset)

Create a new frame and click copy text button and paste the contents into the frame.

Save your program the run the code in frame.

Here is the output that you should get.

KMeans(n_clusters=4)

We guessed four as to the number of clusters of customers:

Customers with high income that spend a lot
Customers with high income that don't spend a lot.
Customers with low income that spend a lot.
Cutomers with low income and low spending

Now create a new frame and key in: print(km_model.cluster_centers_)

Your results should give you the center of each cluster

[[86.53846154 82.12820513]
[87. 18.63157895]
[26.30434783 20.91304348]
[48.26 56.48 ]]

Now will print the labels: 0,1,2,3 for each group.

Key in the following code in a new frame.

print(km_model.labels_)

You should see the following when you run the program.

[2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0]

Now let's graph this information from our model with a scatter plot that will show us visually the individual four clusters.

Here is the code you need. Create a new frame and paste in the following code.

Create an new frame and click copy text button and paste the contents into the frame.

Save your program the run the code in frame.

Here is the output that you should get. Since the colors are randomly assigned, your colors will probably be different.

It seems like there is an additional cluster: the upper-left-hand corner. The one without the centriod or cluster center.

We need to adjust our code to reflect the extra cluster. Add the following code in a new frame.

km_model = KMeans(n_clusters=5)
km_model.fit(dataset)

Add the following code to produce the scatter graph.

Create a new frame and click copy text button and paste the contents into the frame.

Save your program the run the code in frame.

Here is the output that you should get. Since the colors are randomly assigned, your colors will probably be different.

[[26.30434783 20.91304348]
[86.53846154 82.12820513]
[88.2 17.11428571]
[25.72727273 79.36363636]
[55.2962963 49.51851852]]

Now we see five clusters and the centriod values. Remember the centroid value is similar to the mean. Some customers make more and some make less. The same goes for spending score.

Purple cluster 0 represents customers that make $26,000 and have a spending score of 20.
Cyan cluster 1 represents customers that make $86,000 and have a spending score of 82.
Tan cluster 2 represents customers that make $88,000 and have a spending score of 17.
Green cluster3 represents customers that make $79,000 and have a spending score of 25.
Red cluster 4 represent cusomers that earn $55,000 and have a spending score of 49.

Now let's see which customers belong to each group. The code is listed below

Create a new frame and click copy text button and paste the contents into the frame.

Save your program the run the code in frame.

Here is the output that you should get.

data_index	cluster
0	0	0
1	1	3
2	2	0
3	3	3
4	4	0
5	5	3
6	6	0
7	7	3
8	8	0
9	9	3
10	10	0
11	11	3
12	12	0
13	13	3
14	14	0
15	15	3
16	16	0
17	17	3
18	18	0
19	19	3
20	20	0
21	21	3
22	22	0
23	23	3
24	24	0
25	25	3
26	26	0
27	27	3
28	28	0
29	29	3
30	30	0
31	31	3
32	32	0
33	33	3
34	34	0
35	35	3
36	36	0
37	37	3
38	38	0
39	39	3
40	40	0
41	41	3
42	42	0
43	43	4
44	44	0
45	45	3
46	46	4
47	47	4
48	48	4
49	49	4
50	50	4
51	51	4
52	52	4
53	53	4
54	54	4
55	55	4
56	56	4
57	57	4
58	58	4
59	59	4
60	60	4
61	61	4
62	62	4
63	63	4
64	64	4
65	65	4
66	66	4
67	67	4
68	68	4
69	69	4
70	70	4
71	71	4
72	72	4
73	73	4
74	74	4
75	75	4
76	76	4
77	77	4
78	78	4
79	79	4
80	80	4
81	81	4
82	82	4
83	83	4
84	84	4
85	85	4
86	86	4
87	87	4
88	88	4
89	89	4
90	90	4
91	91	4
92	92	4
93	93	4
94	94	4
95	95	4
96	96	4
97	97	4
98	98	4
99	99	4
100	100	4
101	101	4
102	102	4
103	103	4
104	104	4
105	105	4
106	106	4
107	107	4
108	108	4
109	109	4
110	110	4
111	111	4
112	112	4
113	113	4
114	114	4
115	115	4
116	116	4
117	117	4
118	118	4
119	119	4
120	120	4
121	121	4
122	122	4
123	123	1
124	124	2
125	125	1
126	126	4
127	127	1
128	128	2
129	129	1
130	130	2
131	131	1
132	132	4
133	133	1
134	134	2
135	135	1
136	136	2
137	137	1
138	138	2
139	139	1
140	140	2
141	141	1
142	142	4
143	143	1
144	144	2
145	145	1
146	146	2
147	147	1
148	148	2
149	149	1
150	150	2
151	151	1
152	152	2
153	153	1
154	154	2
155	155	1
156	156	2
157	157	1
158	158	2
159	159	1
160	160	2
161	161	1
162	162	2
163	163	1
164	164	2
165	165	1
166	166	2
167	167	1
168	168	2
169	169	1
170	170	2
171	171	1
172	172	2
173	173	1
174	174	2
175	175	1
176	176	2
177	177	1
178	178	2
179	179	1
180	180	2
181	181	1
182	182	2
183	183	1
184	184	2
185	185	1
186	186	2
187	187	1
188	188	2
189	189	1
190	190	2
191	191	1
192	192	2
193	193	1
194	194	2
195	195	1
196	196	2
197	197	1
198	198	2
199	199	1

Day 5: Finding the customers to target

Since we are interested in increasing our sales as any enterprise would, we should concentrate our marketing efforts to target customers with higher incomes and like to spend a lot in our store, This is the cyan cluster 1, the cluster at the top right of our scatter plot. The average income of this group is $88,000 and they have a spending score of 82. Let's see if our model can isolate this cluster. Here is the code.

Create a new frame and click copy text button and paste the contents into the frame.

Save your program the run the code in the frame.

Here is the output that you should get.

There are a total of 35 of our customers that we want to target since they earn and spend the most. If we look at our original file, the Excel one, we can get some more detail about these customers. For example, customer id 124 is a 39 year old male making $69,000 a year and has a spending score of 91 . In mathematics and physics, the centroid, also known as geometric center or center of figure, of a plane figure or solid figure is the arithmetic mean position of all the points in the surface of the figure.²

Since we can isolate individual customers in the targeted cluster, we would want to find our copy of their sales slips to see what they purchased. We can then tailor our marketing efforts to reflect top of the line merchandise for these items and create an email campaign or direct marketing effort for these customers.