(New section: Advantage and Disadvantage of K-NN Algorithm) |
|||
| (One intermediate revision by the same user not shown) | |||
| Line 2: | Line 2: | ||
We will estimate value of Y for x = 6. | We will estimate value of Y for x = 6. | ||
| − | X | + | X = 1 2 3 4 7 6 |
| − | + | Y = 5 9 15 20 30 ? | |
| − | + | ||
| − | + | ||
| − | + | ||
| − | + | ||
| − | + | ||
1. Decide K. In this example let K = 2. | 1. Decide K. In this example let K = 2. | ||
2. Find distance from the current X value | 2. Find distance from the current X value | ||
| + | |||
| + | distance: 5 4 3 2 1 | ||
3. Decide K-NN value => 20 & 30 | 3. Decide K-NN value => 20 & 30 | ||
| Line 21: | Line 18: | ||
Answer: Estimated(Predicted) Y = 25 | Answer: Estimated(Predicted) Y = 25 | ||
| + | |||
| + | == Advantage and Disadvantage of K-NN Algorithm == | ||
| + | |||
| + | |||
| + | Advantage | ||
| + | 1. Strong to noisy data | ||
| + | 2. Works very well for large training data | ||
| + | |||
| + | Disadvantage | ||
| + | 1. Highly dependent on the parameter K | ||
| + | 2. Computational cost is very high since we need to calculate distance for every input from the traing samples | ||
| + | 3. Performance varies depending on distance measure | ||
Latest revision as of 22:10, 5 April 2008
Time Series Estimation Problem We will estimate value of Y for x = 6.
X = 1 2 3 4 7 6 Y = 5 9 15 20 30 ?
1. Decide K. In this example let K = 2.
2. Find distance from the current X value
distance: 5 4 3 2 1
3. Decide K-NN value => 20 & 30
4. Estimate Y value by taking K mean values of X
Y = (20+30)/2 = 25
Answer: Estimated(Predicted) Y = 25
Advantage and Disadvantage of K-NN Algorithm
Advantage 1. Strong to noisy data 2. Works very well for large training data
Disadvantage 1. Highly dependent on the parameter K 2. Computational cost is very high since we need to calculate distance for every input from the traing samples 3. Performance varies depending on distance measure
