stat.py

# -*- coding: utf-8 -*-
from collections import OrderedDict

prices = [
390000,
550000,
550000,
580000,
665000,
670000,
700000,
750000,
755000,
760000,
800000,
830000,
850000,
850000,
870000,
870000,
890000,
899000,
900000,
900000,
900000,
900000,
900000,
900000,
915000,
923574,
930000,
930000,
930000,
930000,
930000,
945000,
949000,
950000,
950000,
950000,
950000,
950000,
950000,
950000,
950000,
950000,
950000,
950000,
950000,
955740,
960000,
960000,
965000,
969000,
        ]

def average(items):
    if not items:
        return 0
    return sum(items) / len(items)

def calc(prices, count_parts):
    _min, _max = min(prices), max(prices)
    step = (_max - _min) / count_parts
    #print step
    return [average(filter(lambda price: i * step <= price < (i + 1) * step, prices))
            for i in range(count_parts)]

from scipy import stats


def average(items):
    if not items:
        return 0
    return sum(items) / len(items)

def calc(prices, step_percent):
    _min, _max = min(prices), max(prices)
    prices = sorted(prices)

    step_limit = (_max - _min) * step_percent / 100
    while 1:
        current_min_range = prices[-1]

        new_prices = []
        for i in range(len(prices) - 1):
            price_range = abs(prices[i] - prices[i + 1])
            if price_range < current_min_range:
                current_min_range = price_range
            new_prices.append(average([prices[i], prices[i + 1]]))
        if current_min_range >= step_limit:
            break
        prices = new_prices
    return prices



def rec_calc(prices, count_parts):
    prices = sorted(prices)

    while len(prices) > count_parts:
        new_prices = []
        ranges = sorted(abs(prices[i] - prices[i + 1])
                          for i in range(len(prices) - 1))
        max_range = ranges[len(ranges) / 2]

        for i in range(len(prices) - 1):
            if abs(prices[i] - prices[i + 1]) < max_range:
                new_prices.append(average([prices[i], prices[i + 1]]))
            else:
                new_prices.append(prices[i])
        prices = new_prices
    return prices

def calc(prices, count_parts):
    prices = sorted(prices)

    while len(prices) > count_parts:
        _min = prices[-1]
        _min_i = None

        for i in range(len(prices) - 1):
            v = abs(prices[i] - prices[i + 1])
            if v < _min:
                _min = v
                _min_i = i
        prices[_min_i:_min_i+2] = [average([prices[_min_i], prices[_min_i + 1]])]
    return prices

#print calc(prices, 10)

_prices = [
    20,
    50,
    51,
    52,
    52,
    70,
    100,
]

def average(items):
    if not items:
        return 0
    return sum(items) / len(items)

def calc(prices, count_parts):
    prices = sorted(prices)
    prices_counts = [1 for i in prices]
    prices_len = len(prices)

    while len(prices) > count_parts:
        _min = prices[-1]
        _min_i = None

        for i in range(len(prices) - 1):
            v = abs(prices[i] - prices[i + 1])
            if v < _min:
                _min = v
                _min_i = i
        prices[_min_i:_min_i + 2] = [average([prices[_min_i], prices[_min_i + 1]])]
        prices_counts[_min_i:_min_i + 2] = [prices_counts[_min_i] + prices_counts[_min_i + 1]]
    return zip(prices, [100 * i / prices_len for i in prices_counts])


#for k, v in calc(prices, 10):
#    print "%s - %s%%" % (k, v)


# percentiles = OrderedDict([ ('p_%d' % x, stats.scoreatpercentile(prices, x*10)) for x in range(11) ])
#
# for x in range(11):
#     percentile = percentiles['p_%d' % x]
#     print stats.percentileofscore(prices, percentile), percentile


prices_list = [
    [
        390000, 550000, 550000, 580000, 665000, 670000, 700000, 750000, 755000, 760000,
        800000, 830000, 850000, 850000, 870000, 870000, 890000, 899000, 900000, 900000,
        900000, 900000, 900000, 900000, 915000, 923574, 930000, 930000, 930000, 930000,
        930000, 945000, 949000, 950000, 950000, 950000, 950000, 950000, 950000, 950000,
        950000, 950000, 950000, 950000, 950000, 955740, 960000, 960000, 965000, 969000,
    ],
    [20, 50, 51, 52, 52, 70, 100],
]


def average(items):
    if not items:
        return 0
    return sum(items) / len(items)


def get_min_range(prices):
    return min(abs(prices[i] - prices[i + 1])
               for i in range(len(prices) - 1))


def calc(prices, max_parts):
    prices = sorted(prices)
    prices_counts = [1 for i in prices]
    prices_len = len(prices)
    min_range = sorted(abs(prices[i] - prices[i + 1])
                       for i in range(len(prices) - 1))[len(prices) / 2]

    while len(prices) > 2 and (len(prices) > max_parts or get_min_range(prices) < min_range):
        _min = prices[-1]
        _min_i = None

        for i in range(len(prices) - 1):
            v = abs(prices[i] - prices[i + 1])
            if v < _min:
                _min = v
                _min_i = i
        prices[_min_i:_min_i + 2] = [average([prices[_min_i], prices[_min_i + 1]])]
        prices_counts[_min_i:_min_i + 2] = [prices_counts[_min_i] + prices_counts[_min_i + 1]]
    return zip(prices, prices_counts)
    return zip(prices, [100 * i / prices_len for i in prices_counts])


for prices in prices_list:
    print '=' * 78
    print '-' * 78
    print prices
    print '-' * 78

    all = 0
    for k, v in calc(prices, 10):
        print "%s - %s" % (k, v)
        all+=v
    assert all == len(prices)



#print prices

percentiles = OrderedDict([ ('p_%d' % x, stats.scoreatpercentile(prices, x*10)) for x in range(11) ])

for x in range(11):
    percentile = percentiles['p_%d' % x]
    #print stats.percentileofscore(prices, percentile), percentile

prices_count = float(len(prices))

all = 0

# #print [(x, x*5) for x in range(21) if x%2]
# for i in [ x*5 for x in range(21) if x%2]:
#     start = i-5
#     stop = i+5
#     start_p = stats.scoreatpercentile(prices, start)
#     stop_p = stats.scoreatpercentile(prices, stop)
#     count = len([x for x in prices if x >= start_p and x <=stop_p])
#     perc = count/prices_count*100 if count else 0
#     print i, start_p, stop_p, count, perc
#     all+=count
#
# print prices_count, all