In [1]:

    

import os 
#os.chdir("/home/psenin/git/saxpy")
os.chdir('C:/Users/ng6e164/PycharmProjects/saxpy')
os.getcwd()


    

    
Out[1]:





'C:\\Users\\ng6e164\\PycharmProjects\\saxpy'

In [2]:

    

import numpy as np
from saxpy.sax import sax_via_window
from saxpy.saxvsm import series_to_wordbag
from saxpy.saxvsm import manyseries_to_wordbag
from saxpy.saxvsm import bags_to_tfidf
from saxpy.saxvsm import cosine_similarity
from saxpy.saxvsm import class_for_bag


    

In [ ]:

    




    

In [3]:

    

bag1 = { "this": 10, "is": 1, "a": 2, "sample": 1}
bag2 = { "this": 10, "is": 1, "another": 2, "example": 3}
bags = {"bag1":bag1, "bag2":bag2}
str(bags)


    

    
Out[3]:





"{'bag1': {'this': 10, 'is': 1, 'a': 2, 'sample': 1}, 'bag2': {'this': 10, 'is': 1, 'another': 2, 'example': 3}}"

In [4]:

    

[*bags.copy()]


    

    
Out[4]:





['bag1', 'bag2']

In [5]:

    

res = bags_to_tfidf(bags)
res


    

    
Out[5]:





{'classes': ['bag1', 'bag2'],
 'vectors': {'a': [0.76150001041880899, 0.0],
  'another': [0.0, 0.76150001041880899],
  'example': [0.0, 0.96090602783640278],
  'sample': [0.48045301391820139, 0.0]}}

In [6]:

    

res["vectors"]


    

    
Out[6]:





{'a': [0.76150001041880899, 0.0],
 'another': [0.0, 0.76150001041880899],
 'example': [0.0, 0.96090602783640278],
 'sample': [0.48045301391820139, 0.0]}

In [7]:

    

from saxpy.util import read_ucr_data
dd = read_ucr_data('resources/data/cbf/CBF_TRAIN')


    

In [8]:

    

win = 30
paa = 6
alp = 6
na_strategy = "exact"
ztresh = 0.01

bags = {}

for key, arr in dd.items():
    print(key)
    bags[key] = manyseries_to_wordbag(dd[key], win, paa, alp, na_strategy, ztresh)


    

    




1
3
2

In [9]:

    

[*bags.copy()]


    

    
Out[9]:





['1', '3', '2']

In [10]:

    

vectors = bags_to_tfidf(bags)
vectors['classes']


    

    
Out[10]:





['1', '3', '2']

In [ ]:

    




    

In [ ]:

    




    

In [11]:

    

dt = read_ucr_data('resources/data/cbf/CBF_TEST')


    

In [12]:

    

import matplotlib.pyplot as plt
import datetime
import numpy as np

series = dt['1'][6]
x = np.arange(0, len(series))
y = np.asarray(series)

plt.plot(x,y)
plt.show()


    

In [13]:

    

test_bag = series_to_wordbag(series, 30, 6, 6, "exact", 0.01)
#test_bag


    

In [14]:

    

res = cosine_similarity(vectors, test_bag)
res


    

    
Out[14]:





{'1': 0.8096998752732335, '2': 0.91433733786773486, '3': 0.95867368251031904}

In [15]:

    

class_for_bag(res)


    

    
Out[15]:





'1'

In [16]:

    

for cls in [*dt.copy()]:
    print(cls)
    i = 0
    for s in dt[cls]:
        sim = cosine_similarity(vectors, 
                                series_to_wordbag(s, 30, 6, 6, "none", 0.01))
        res = class_for_bag(sim)
        if res != cls:
            print(" misclassified", i, "as", res, sim)
        i = i + 1


    

    




2
1
 misclassified 72 as 2 {'1': 0.92853215910979048, '3': 0.94644982977028103, '2': 0.90361918327181945}
 misclassified 91 as 2 {'1': 0.89773989722580827, '3': 0.95640189691324418, '2': 0.89753754493831839}
 misclassified 256 as 2 {'1': 0.93156990542330831, '3': 0.94594144806166813, '2': 0.90273991600777048}
3
 misclassified 223 as 1 {'1': 0.86883795152559462, '3': 0.89417006609748306, '2': 0.98599573407109053}

In [17]:

    

series = dt['1'][72]
x = np.arange(0, len(series))
y = np.asarray(series)
plt.plot(x,y)
plt.show()


    

In [18]:

    

series = dt['3'][223]
x = np.arange(0, len(series))
y = np.asarray(series)
plt.plot(x,y)
plt.show()


    

In [20]:

    

train = read_ucr_data('resources/data/cbf/CBF_TRAIN')
test = read_ucr_data('resources/data/cbf/CBF_TEST')

win = 30
paa = 6
alp = 6
na_strategy = "exact"
zthresh = 0.01

def test_accuracy(dd_train, dd_test, sax_win, sax_paa, sax_alp, sax_strategy, z_threshold):
    
    train_bags = {}
    for key, arr in dd_train.items():
        train_bags[key] = manyseries_to_wordbag(dd[key], sax_win, sax_paa,
                                                sax_alp, sax_strategy, z_threshold)
    
    tfidf_vectors = bags_to_tfidf(train_bags)

    correct = 0
    count = 0

    for cls in [*dd_test.copy()]:
        for s in dd_test[cls]:
            sim = cosine_similarity(tfidf_vectors, 
                                    series_to_wordbag(s, sax_win, sax_paa,
                                                      sax_alp, sax_strategy, z_threshold))
            res = class_for_bag(sim)
            if res == cls:
                correct = correct + 1
            count = count + 1
    
    return correct / count


    

In [21]:

    

test_accuracy(train, test, win, paa, alp, na_strategy, zthresh)


    

    
Out[21]:





0.9944444444444445

In [ ]:

    




    

In [ ]: