Please enable JavaScript.
Coggle requires JavaScript to display documents.
Machine Learning (Application (Fraud Detection, Better Marketing, Market…
Machine Learning
Data
Mining
-
-
-
Why Important?
Large Amount of Data
- Accumulated data (30 years+)
- Collect information for analysis
Affordable Computing Power
- Reduction price of computer systems
- Data mining is computationally expensive
Evolution of Technology
- Improved database
- Fast & cheap data collection
- Ability to analyze & synthesize information
Statistical & Learning Algorithm
- Enable development of new algorithm
- Adapted algorithm from AI
Strong Business Competition
- Growth in service economies
- Information rich & competitive
Statistical
vs ML
Statistical
- Confirm hypothesis
- Data requirement
- More assumptions
- Design importance
- General Population Prediction
ML
- Generate hypothesis
- Exploratory
- Less Assumption
- Less Data Preparation
- Result Oriented
-
Supervised
Profile & Predict
- Build predictive profile of historical outcome
- Utilize collection of potential inputs
- Supervise the process as the algorithm
attempt to model outcome
- Explore all possible combination & interaction
- Use that profile to predict future cases
Neural Network
- Predict outcome based on input
- Input are weighted on hidden layers
- Require minimal statistical or maths knowledge
- Back-propagate to adjust weights
Decision Tree
- Excellent model in complex relationship
- Easy understand
- High accuracy in small data
-
Unsupervised
Clustering
- Reveal natural groups within a dataset
- No prior knowledge about group/characteristics
- Exploratory data analysis technique
Association
- Find things occurs together
- Can exist between any of attributes
Sequential Association
- Discover association rules in
time-oriented data
- Find sequence/order of events
-
-