-
Notifications
You must be signed in to change notification settings - Fork 0
/
part12.py
177 lines (141 loc) · 4.93 KB
/
part12.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
# -*- coding: utf-8 -*-
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
np.random.seed(12345)
np.set_printoptions(precision=2,suppress=True)
def auto_regression(d_tau, sales_data, test_data=None):
sales_data_true = sales_data.copy()
N_item = sales_data.shape[0]
nan_locations = np.nonzero(np.isnan(sales_data))
sales_data[nan_locations] = np.random.rand() + 300
for i in range(30):
# Phase 1: solve for weights
b, A_index, A = setup_phase_1(d_tau, sales_data)
w = np.linalg.lstsq(A, b)[0]
if len(nan_locations[0]) == 0: break
# # Phase 2: solve for nans
b, A = setup_phase_2(d_tau, b, A_index, A, w, sales_data_true)
k = np.linalg.lstsq(A, b)[0]
sales_data[nan_locations] = k
rv = np.reshape(w, [N_item, N_item, d_tau])
return rv
def setup_phase_2(d_tau, b, A_index, A, w ,sales_data):
N_row = len(b)
nan_locations = np.nonzero(np.isnan(sales_data))
b_new = []
A_new = []
for i in range(N_row):
include = check_row(i,d_tau, b, A, A_index, sales_data, w)
if include:
b_new.append(include[0])
A_new.append(include[1])
return b_new, A_new
def check_row(i, d_tau, b, A, A_index, sales_data, w):
N_item = sales_data.shape[0]
nan_locations = np.nonzero(np.isnan(sales_data))
nan_tuples = np.vstack([nan_locations[0], nan_locations[1]])
nan_tuples = to_tuple(nan_tuples)
N_nans = len(nan_tuples)
bi = calculate_b_index(d_tau, N_item, i)
to_check = np.vstack([bi, A_index[i,:]])
c = 0
qualified = False
current_b = 0
current_row = [0] * N_nans
for elem in to_check:
elem = tuple(elem)
if elem != (-1,-1):
if elem in nan_tuples:
qualified = True
index = nan_tuples.index(elem)
if c == 0:
current_row[index] = -1
else:
current_row[index] = w[c-1]
else:
if c > 0:
current_b += (-w[c-1]) * sales_data[elem]
else:
current_b += b[i]
c += 1
if qualified:
return current_b, current_row
else:
return False
def to_tuple(v):
rv = []
for el in v.T:
rv.append((el[0], el[1]))
return rv
def setup_phase_1(d_tau, sales_data):
N_item = sales_data.shape[0]
N_window = d_tau * N_item
row_len = d_tau * (N_item ** 2)
b = sales_data[:,d_tau:].flatten(order='F') # vectorize column by column
col_len = len(b)
A = np.zeros([col_len ,row_len])
A_index = np.zeros([col_len ,row_len], dtype=(int, 2)) - 1
for i in range(col_len): # iterate for each row
item, month = calculate_b_index(d_tau, N_item, i)
r = np.arange(0, N_item)
c = np.arange(month - d_tau, month)
row_index, r, c = form_tuple_list(r, c)
row = sales_data[r, c]
start = item * N_window
end = start + N_window
A[i,start:end] = row
A_index[i,start:end] = row_index
return b, A_index, A
def calculate_b_index(d_tau, N_item, index):
month = index / N_item + d_tau
item = index % N_item
return item, month
def form_tuple_list(a1, a2):
rv1= []
rv2= []
rv3= []
for e1 in a1:
for e2 in a2:
rv1.append((e1,e2))
rv2.append(e1)
rv3.append(e2)
return rv1, rv2, rv3
def next_six_months(d_tau, w, sales_data):
N_item = sales_data.shape[0]
for i in range(6):
future_sales = np.zeros([N_item, 1])
for j in range(N_item):
a = sales_data[:, -d_tau:]
wj = w[j,:,:]
future_sales[j] = np.sum(a * wj)
sales_data = np.hstack([sales_data, future_sales])
return sales_data[:,-6:]
def remove_nans(row_vector):
months = np.nonzero(~np.isnan(row_vector))[0]
sales = row_vector[months]
return months, sales
def project_part1_2(d_tau, train, test=None):
trc = train.copy()
w = auto_regression(d_tau, trc)
future_months = np.arange(trc.shape[1],trc.shape[1]+6)
future_sales = next_six_months(d_tau, w, trc)
errors = None
if test is not None:
errors = test - future_sales
plt.figure()
colors = ['r','k','g','b','m','y','c']
for i in range(train.shape[0]):
past_months, past_sales = remove_nans(train[i, :])
plt.plot(past_months, past_sales, '--o', color=colors[i],lw=0.7, ms=6.0)
# missing = np.nonzero(np.isnan(R_true[i,:]))
# plt.plot(missing ,train[i, missing], 's', color=colors[i], ms=8)
plt.plot(future_months, future_sales[i,:], '^', color=colors[i], ms=8)
plt.show()
return w, future_months, future_sales, errors
R_true = np.genfromtxt('dataset12.csv', delimiter=',',dtype=float)
train = R_true.copy()
test = None
# train = R_true[:,:18].copy()
# test = R_true[:,18:].copy()
w, pred_month, pred_sales, errors = project_part1_2(2, train, test)