Analyzing text with Nexa

This is an analysis of the text from the financial times with the nexa framework. Here we apply the nexa machinery to it



In [1]:

    
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import h5py
import IPython

import sys
sys.path.append('../')

from inputs.sensors import Sensor, PerceptualSpace
from inputs.lag_structure import LagStructure
from nexa.nexa import Nexa# First we have to load the signal









    



/home/heberto/miniconda/envs/nexa/lib/python3.5/site-packages/sklearn/utils/fixes.py:64: DeprecationWarning: inspect.getargspec() is deprecated, use inspect.signature() instead
  if 'order' in inspect.getargspec(np.copy)[0]:

Extract the Data

Now we extract the signal



In [2]:

    
signal_location = '../data/wall_street_data_small.hdf5'

# Access the data and load it into signal
with h5py.File(signal_location, 'r') as f:
    dset = f['signal']
    signals = np.empty(dset.shape, np.float)
    dset.read_direct(signals)

Reshape the data for our purposes and take a piece of it



In [3]:

    
# Reshape the data and limit it
Ndata = 10000
signals = signals.reshape(signals.shape[0], signals.shape[1] * signals.shape[2])
# signals = signals[:Ndata, ...].astype('float')
signals += np.random.uniform(size=signals.shape)
print('zeros', np.sum(signals[0] == 0))
print('signals shape', signals.shape)









    



zeros 0
signals shape (544270, 400)

Perceptual Space



In [4]:

    
dt = 1.0
lag_times = np.arange(0, 10, 1)
window_size = signals.shape[0] - (lag_times[-1] + 1)
weights = None

lag_structure = LagStructure(lag_times=lag_times, weights=weights, window_size=window_size)
sensors = [Sensor(signal, dt, lag_structure) for signal in signals.T]
perceptual_space = PerceptualSpace(sensors, lag_first=True)

Nexa Machinery



In [5]:

    
# Get the nexa machinery right
Nspatial_clusters = 3
Ntime_clusters = 4
Nembedding = 2

nexa_object = Nexa(perceptual_space, Nspatial_clusters, Ntime_clusters, Nembedding)



In [ ]:

    
# Now we calculate the distance matrix
nexa_object.calculate_distance_matrix()
nexa_object.calculate_embedding()



In [11]:

    
# Now we calculate the clustering
nexa_object.calculate_spatial_clustering()



In [17]:

    
# We calculate the cluster to index
nexa_object.calculate_cluster_to_indexes()



In [ ]:

    
# Data clusters
nexa_object.calculate_time_clusters()



In [ ]: