Introduction to Data Mining

Lecture 1

Dr. Greg Chism

University of Arizona
INFO 523 - Spring 2024

Warm up

Announcements

  • Reading Quiz #1 is due Friday, Jan 19th, 11:59pm

  • Project 1 overview next week

    • Teams will be announced as well

What is data mining?

One of many definitions:

“Data mining is the science of extracting useful knowledge from huge data repositories.” - ACM SIGKDD, Data Mining Curriculum: A Proposal

What is data mining?

Convergence of several fields

  • Statistics

  • Computer science (machine learning, AI)

  • Data science

  • Optimization

Why data mining?
Commercial viewpoint

Businesses collect + store tons of data

  • Purchases at department/grocery stores
  • Bank/credit card transactions
  • Web and social media data
  • Mobile and IOT

Computers are cheaper + more powerful

Competition to provide better services

  • Mass customization and recommendation systems
  • Targeted advertising
  • Improved logistics

Knowledge discovery in databases (KDD)

Data mining tasks

Descriptive

Find human-interpretable patterns that describes the data


Predictive Methods

Use features to predict unknown or future values of another feature

Predictive modeling

Classification

Budget Duration Channel Target_Audience_Size Season Campaign_Success
0 1.170199 -1.110199 1 -0.910081 1 0
1 0.198680 1.499221 0 0.988123 1 1
2 1.695723 0.126734 0 0.222349 0 1
3 -0.531455 -0.169270 1 -2.224787 0 0
4 0.194052 -2.479343 0 1.518522 2 0

Regression

Feature Target
0 0.931280 50.779929
1 0.087047 -10.065270
2 -1.057711 -34.918392
3 0.314247 10.526743
4 -0.479174 -17.738377

Classification

Find a model for the class attribute as a function of the other attributes

Goal: assign new records to a class as accurately as possible.

E.g., Customer Attrition, Directed Marketing

Budget Duration Channel Target_Audience_Size Season Campaign_Success
0 1.170199 -1.110199 1 -0.910081 1 0
1 0.198680 1.499221 0 0.988123 1 1
2 1.695723 0.126734 0 0.222349 0 1
3 -0.531455 -0.169270 1 -2.224787 0 0
4 0.194052 -2.479343 0 1.518522 2 0
5 1.807197 1.341938 1 -0.667830 1 1
6 -0.093387 -2.407374 0 -0.480568 2 0
7 1.063941 0.864310 1 -0.957300 1 1
8 -1.433152 1.360601 1 1.384636 1 1
9 -0.937926 0.464292 1 -2.113015 0 1

Regression

Find a model that predicts a variable (Y) from another variable (X)

Both are continuous variables (floats)

Feature Target
0 0.931280 50.779929
1 0.087047 -10.065270
2 -1.057711 -34.918392
3 0.314247 10.526743
4 -0.479174 -17.738377
5 0.647689 31.564596
6 -0.463418 -30.068883
7 0.542560 5.912007
8 0.611676 23.473374
9 1.003533 32.343595

Association mining

Given a set of transactions, produce rules of association

  antecedents consequents  antecedent support  consequent support  support  \
0   (Diapers)      (Beer)                 0.8                 0.6      0.6   
1      (Beer)   (Diapers)                 0.6                 0.8      0.6   
2     (Bread)   (Diapers)                 0.8                 0.8      0.6   
3   (Diapers)     (Bread)                 0.8                 0.8      0.6   
4     (Bread)      (Milk)                 0.8                 0.8      0.6   
5      (Milk)     (Bread)                 0.8                 0.8      0.6   
6      (Milk)   (Diapers)                 0.8                 0.8      0.6   
7   (Diapers)      (Milk)                 0.8                 0.8      0.6   

   confidence    lift  leverage  conviction  zhangs_metric  
0        0.75  1.2500      0.12         1.6           1.00  
1        1.00  1.2500      0.12         inf           0.50  
2        0.75  0.9375     -0.04         0.8          -0.25  
3        0.75  0.9375     -0.04         0.8          -0.25  
4        0.75  0.9375     -0.04         0.8          -0.25  
5        0.75  0.9375     -0.04         0.8          -0.25  
6        0.75  0.9375     -0.04         0.8          -0.25  
7        0.75  0.9375     -0.04         0.8          -0.25

Association mining

  • Let the rule discovered be: {Potato Chips, …} → {Soft drink}

  • Soft drink as RHS: what can boost sales? Discount Potato Chips?

  • Potato Chips as LHS: which products are affected if Potato Chips are discontinued

  • Potato Chips in LHS and Soft drink in RHS: What products should be sold with Potato Chips to promote sales of Soft drinks!

Association mining goals

  • Goal: Anticipate the nature of repairs to keep the service vehicles equipped with right parts to speed up repair time.
  • Approach: Process the data on tools and parts required in previous repairs at different consumer locations and discover co-occurrence patterns.

Clustering

  • Group points that are similar to one another

  • Separate dissimilar points

  • Groups are not known → Unsupervised Learning

  • E.g., Market Segmentation, Document Types

Anomaly detection

Detect significant deviations from normal behavior.

Other data mining tasks

Challenges of data mining

Conclusions

Data Mining is interdisciplinary

  • Statistics

  • CS (machine learning, AI)

  • Data science

  • Optimization

Data mining is a team effort

  • Data management

  • Statistics

  • Programming

  • Communication

  • Application domain

Intro to NumPy

What is NumPy?

  • NumPy = Numerical Python

  • Foundational package for scientific computing

  • High-performance multidimensional arrays

  • Tools for working with arrays

Start with

pip install numpy

Why NumPy for data mining?

  • Essential for data processing, manipulation, and analysis.

  • Underpins advanced data mining algorithms implemented in Python.

  • Fast and memory-efficient with powerful data structures.

NumPy Arrays

Creating arrays

import numpy as np

# Creating a simple NumPy array
arr = np.array([1, 2, 3, 4])

# Multidimensional array
multi_arr = np.array([[1, 2, 3], [4, 5, 6]])

# Range of values
range_arr = np.arange(10)

# Array of zeros
zeros_arr = np.zeros((3, 3))

# Array of ones
ones_arr = np.ones((2, 2))

# Identity matrix
identity_matrix = np.eye(3)
arr: [1 2 3 4] 

multi_arr: [[1 2 3]
 [4 5 6]] 

range_arr: [0 1 2 3 4 5 6 7 8 9] 

zeros_arr: [[0. 0. 0.]
 [0. 0. 0.]
 [0. 0. 0.]] 

ones_arr: [[1. 1.]
 [1. 1.]] 

identity_matrix: [[1. 0. 0.]
 [0. 1. 0.]
 [0. 0. 1.]]

Array attributes

# Array dimensions
print("Dimensions:", multi_arr.ndim)

# Shape of array
print("Shape:", multi_arr.shape)

# Size of array
print("Size:", multi_arr.size)

# Data type of array elements
print("Data Type:", multi_arr.dtype)
Dimensions: 2
Shape: (2, 3)
Size: 6
Data Type: int64

Array operations

Arithmetic operations

# Element-wise addition
addition = arr + 2

# Element-wise subtraction
subtraction = arr - 2

# Element-wise multiplication
multiplication = arr * 2

# Element-wise division
division = arr / 2
addition: [3 4 5 6] 

subtraction: [-1  0  1  2] 

multiplication: [2 4 6 8] 

division: [0.5 1.  1.5 2. ]

Aside

Why do my outputs look than different than Python?

print(addition)
print(subtraction)
print(multiplication)
print(division)
[3 4 5 6]
[-1  0  1  2]
[2 4 6 8]
[0.5 1.  1.5 2. ]
print("addition:", addition, "\n")

print("subtraction:", subtraction, "\n")

print("multiplication:", multiplication, "\n")

print("division:", division, "\n")
addition: [3 4 5 6] 

subtraction: [-1  0  1  2] 

multiplication: [2 4 6 8] 

division: [0.5 1.  1.5 2. ]

Statistical operations

# Sum of elements
total = arr.sum()

# Mean of elements
mean_value = arr.mean()

# Standard deviation
std_dev = arr.std()

# Correlation coefficient
corr = np.corrcoef(multi_arr)
total: 10 

mean_value: 2.5 

std_dev: 1.118033988749895 

corr: [[1. 1.]
 [1. 1.]]

Advanced operations

Reshaping and transposing

# Reshaping an array
reshaped = np.reshape(range_arr, (2, 5))

# Transpose of an array
transposed = multi_arr.T
range_arr: [0 1 2 3 4 5 6 7 8 9] 

reshaped: [[0 1 2 3 4]
 [5 6 7 8 9]] 

multi_arr: [[1 2 3]
 [4 5 6]] 

transposed: [[1 4]
 [2 5]
 [3 6]]

Indexing and slicing

# Accessing a specific element
element = multi_arr[0, 1]

# Slicing a row
row = multi_arr[1, :]

# Slicing a column
column = multi_arr[:, 2]
multi_arr: [[1 2 3]
 [4 5 6]] 

row: [4 5 6] 

column: [3 6]

Broadcasting

# Broadcasting allows arithmetic operations on arrays of different sizes
broadcasted_addition = multi_arr + np.array([1, 0, 1])
multi_arr: [[1 2 3]
 [4 5 6]] 

broadcasted_addition: [[2 2 4]
 [5 5 7]]

Linear algebra in NumPy

Matrix operations

Dot product: take two equal-length sequences and return a single number

2 • (1, 2, 3) = 2x1 = 2; 2x2 = 0; 2x3 = 6

Matrix multiplication:

(1, 2, 3) • (7, 9, 11) = (1×7 + 2×9 + 3×11) = 58

# Dot product
dot_product = np.dot(arr, arr)

# Matrix multiplication
matrix_mul = np.dot(multi_arr, identity_matrix)
dot_product: 30 

matrix_mul: [[1. 2. 3.]
 [4. 5. 6.]]

Eigenvalues and Eigenvectors

# Eigenvalues and eigenvectors
eigenvalues, eigenvectors = np.linalg.eig(identity_matrix)
eigenvalues: [1. 1. 1.] 

eigenvectors: [[1. 0. 0.]
 [0. 1. 0.]
 [0. 0. 1.]]

NumPy for data mining

Application in Algorithms

  • NumPy arrays are used in various data mining algorithms like clustering, classification, and neural networks.

Performance

  • NumPy operations are implemented in C, which makes them much faster than standard Python.

Conclusion

  • NumPy is integral to data mining and analysis in Python.

  • It provides efficient and fast operations for array and matrix manipulation.

  • Understanding NumPy is crucial for implementing and customizing data mining algorithms.

🔗 datamineaz.org