hastic-server/analytics/models/general_model.py

from models import Model, AnalyticUnitCache

import utils
import numpy as np
import pandas as pd
import scipy.signal
from scipy.fftpack import fft
from scipy.signal import argrelextrema
import math
from scipy.stats import gaussian_kde
from scipy.stats import norm
from typing import Optional


WINDOW_SIZE = 150


class GeneralModel(Model):

    def __init__(self):
        super()
        self.segments = []
        self.ipats = []
        self.state = {
            'convolve_max': WINDOW_SIZE,
        }
        self.all_conv = []

    def fit(self, dataframe: pd.DataFrame, segments: list, cache: Optional[AnalyticUnitCache]) -> AnalyticUnitCache:
        if type(cache) is AnalyticUnitCache:
            self.state = cache
        self.segments = segments

        data = dataframe['value']
        convolve_list = []
        for segment in segments:
            if segment['labeled']:
                segment_from_index = utils.timestamp_to_index(dataframe, pd.to_datetime(segment['from'], unit='ms'))
                segment_to_index = utils.timestamp_to_index(dataframe, pd.to_datetime(segment['to'], unit='ms'))

                segment_data = data[segment_from_index: segment_to_index + 1]
                if len(segment_data) == 0:
                    continue
                x = segment_from_index + int((segment_to_index - segment_from_index) / 2)
                self.ipats.append(x)
                segment_data = data[x - WINDOW_SIZE : x + WINDOW_SIZE]
                segment_min = min(segment_data)
                segment_data = segment_data - segment_min
                convolve = scipy.signal.fftconvolve(segment_data, segment_data)
                convolve_list.append(max(convolve))

        if len(convolve_list) > 0:
            self.state['convolve_max'] = float(max(convolve_list))
        else:
            self.state['convolve_max'] = WINDOW_SIZE / 3

        return self.state

    def do_predict(self, dataframe: pd.DataFrame) -> list:
        data = dataframe['value']
        pat_data = data[self.ipats[0] - WINDOW_SIZE: self.ipats[0] + WINDOW_SIZE]
        x = min(pat_data)
        pat_data = pat_data - x
        y = max(pat_data)

        for i in range(WINDOW_SIZE * 2, len(data)):
            watch_data = data[i - WINDOW_SIZE * 2: i]
            w = min(watch_data)
            watch_data = watch_data - w
            conv = scipy.signal.fftconvolve(pat_data, watch_data)
            self.all_conv.append(max(conv))
        all_conv_peaks = utils.peak_finder(self.all_conv, WINDOW_SIZE * 2)

        filtered = self.__filter_prediction(all_conv_peaks, data)
        filtered = set(item + WINDOW_SIZE for item in filtered)
        # TODO: convert from ns to ms more proper way (not dividing by 10^6)
        return [(dataframe['timestamp'][x - 1].value / 1000000, dataframe['timestamp'][x + 1].value / 1000000) for x in filtered]

    def __filter_prediction(self, segments: list, data: list):
        if len(segments) == 0 or len(self.ipats) == 0:
            return []
        delete_list = []

        for val in segments:
            if self.all_conv[val] < self.state['convolve_max'] * 0.8:
                delete_list.append(val)

        for item in delete_list:
            segments.remove(item)

        return set(segments)