Time series pattern recognition using generated python code
Generated python code analysing time series and extracting patterns from it.
The generated functions are based on a model created by some researchers (see reseachers).
The model uses accumulators to extract aggregations on features of a pattern occurrences in a time series.
All function generated are based on a aggregator_feature_pattern model (see models).
Summary
Models
Patterns
| Pattern name | Regex |
|---|---|
| bump_on_decreasing_sequence | >><>> |
| decreasing | > |
| increasing | < |
| decreasing_sequence | >(> |
| decreasing_terrace | >=+> |
| dip_on_increasing_sequence | <<><< |
| gorge | (>|(>(=|>)>))(<|(<(=|<)<)) |
| increasing_sequence | <(<|=)*<|< |
| increasing_terrace | <=+< |
| inflexion | <(<|=)>|>(>|=)< |
| peak | <(=|<)(>|=)> |
| plain | >=*< |
| plateau | <=*> |
| proper_plain | >=+< |
| proper_plateau | <=+> |
| steady | = |
| steady_sequence | =+ |
| strictly_increasing_sequence | <+ |
| strictly_decreasing_sequence | >+ |
| summit | (<|(<(=|<)<))(>|(>(=|>)>)) |
| valley | >(=|>)(<|=)< |
| zigzag | (<>)+(<|<>)|(><)+(>|><) |
Example :
Data : [4,4,2,2,3,5,5,6,3,1,1,2,2,2,2,2,2,1]
Signature : =>=<<=<>>=<=====>
Peak regex : <(=|<)*(>|=)*>
Peak pattern occurrences : <<=<>> & <=====>
Features
| Name | Meaning |
|---|---|
| one | Identity feature |
| width | Width of a pattern occurrence |
| height | Height of a pattern occurrence |
| max | Largest value among items of a pattern occurrence |
| min | Lowest value among items of a pattern occurrence |
| surface | Surface of a pattern occurrence |
Aggregators
| Name | Meaning |
|---|---|
| max | Largest value among features of pattern occurrences |
| min | Lowest value among features of pattern occurrences |
| sum | Sum of feature values of pattern occurrences |
How to
How to generate functions
From /src execute the following command :
python generator.py
This should generate generated.py file in /src/generated.
How to test
From /src execute the following command :
python test.py
This should output test results in command window.
Tested functions : 123 out of 396 generated
How to use
Just pick up the desired function from /src/generated/generated.py.
All generated functions are standalone. It means you can just copy-paste them and use it. But, don't forget to import operator in your file as the function operator.add() is used.
Data inputs are expected to be an integer array :
data = [0,1,2,3,4,5,6,7,8,9]Here is an example of a python file using only min_min_increasing_sequence function :
import operator
data = [4,3,5,5,2,1,1,3,3,4,6,6,3,1,3,3]
def min_min_increasing_sequence(data):
C = float('inf')
D = float('inf')
R = float('inf')
currentState = 's'
for i in xrange(1,len(data)):
if(i < len(data)):
C_temp = C
D_temp = D
R_temp = R
if data[i] > data[i-1]:
if currentState == 's':
C = min(min(D_temp,data[i-1]),data[i])
D = float('inf')
currentState = 't'
elif currentState == 't':
C = min(C_temp,min(D_temp,data[i]))
D = float('inf')
currentState = 't'
elif data[i] < data[i-1]:
if currentState == 's':
currentState = 's'
elif currentState == 't':
C = float('inf')
D = float('inf')
R = min(R_temp,C_temp)
currentState = 's'
elif data[i] == data[i-1]:
if currentState == 's':
currentState = 's'
elif currentState == 't':
D = min(D_temp,data[i])
currentState = 't'
return min(R,C)
print min_min_increasing_sequence(data)
# supposed to print '1' Contributors
- Florine Cercle
- Denis Allard