def analysis_hour(self):
# self.record_result('<strong style="color: black; font-size: 24px;">正在分析该商品不同省份的购买量...</strong>')
obj = self.data_frame['creation_time']
obj = obj.dt.hour
obj = obj.value_counts()
obj = obj.sort_index()
index = np.arange(0, 24)
obj = obj.reindex(index, method='ffill', fill_value=0)
ax = obj.plot(xticks=index, kind='line', style='o-', label='评论数量')
obj = self.data_frame['reference_time']
obj = obj.dt.hour
obj = obj.value_counts()
obj = obj.sort_index()
obj = obj.reindex(index, method='ffill', fill_value=0)
obj.plot(xticks=index, kind='line', style='o-', label='购买数量')
ax.set_ylabel('数量')
ax.set_xlabel('24 小时分布')
plt.title('购买/评论该商品的 24 小时时间分布图')
plt.tight_layout()
plt.legend()
filename = '%s_creation_time.png' % self.product_id
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.clf()
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')
def analysis_buy_days(self):
# self.record_result('<strong style="color: black; font-size: 24px;">正在分析该商品不同省份的购买量...</strong>')
obj = self.data_frame['days']
obj = obj.value_counts()
obj = obj.sort_index()
# 如果有超过 20 天后评论的，则合并在一起
if len(obj.index) > 20:
value = obj[obj.index >= 20].sum()
obj = obj.drop(obj[obj.index > 20].index)
obj.values[-1] += value
ax = obj.plot(kind='line', style='ro-', xticks=obj.index)
obj = obj.rename({obj.index[-1]: str(obj.index[-1]) + '+'})
ax.set_xticklabels(labels=obj.index, rotation=0)
count = obj.sum()
for i, val in enumerate(obj.values):
if i <= 5:
ax.text(i - 0.4, val, '%.3f%%' % (val * 1.0 / count * 100))
if i == len(obj.index) - 1:
ax.text(i - 0.4, val, '%.3f%%' % (val * 1.0 / count * 100))
plt.title('该商品用户购买后写下评论的时间关系图')
ax.set_xlabel('写评论时间（天）')
ax.set_ylabel('数量')
plt.tight_layout()
filename = '%s_days.png' % self.product_id
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.clf()
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')
def analysis_user_level(self):
# self.record_result('<strong style="color: black; font-size: 24px;">正在分析该商品不同省份的购买量...</strong>')
obj = self.data_frame['user_level_name']
obj = obj.value_counts()
ax = obj.plot(kind='bar', alpha=self.opacity, color=self.color)
ax.set_xticklabels(obj.index,
rotation=0 if len(obj.index) <= 6 else 45)
count = obj.sum()
for i, val in enumerate(obj.values):
ax.text(i - 0.25, val, '%.2f%%' % (val * 1.0 / count * 100))
plt.title('购买该商品的用户等级分布图')
ax.set_xlabel('用户等级')
ax.set_ylabel('数量')
plt.tight_layout()
filename = '%s_user_level.png' % self.product_id
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.clf()
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')
def analysis_mobile(self):
# self.record_result('<strong style="color: black; font-size: 24px;">正在分析该商品不同省份的购买量...</strong>')
fig_size = plt.rcParams["figure.figsize"]
plt.figure(figsize=(2.4, 2.4))
obj = self.data_frame['is_mobile']
obj = obj.value_counts()
obj = obj.rename({1: '移动端', 0: 'PC'})
plt.pie(x=obj.values,
autopct='%.0f%%',
radius=0.7,
labels=obj.index,
startangle=180)
plt.title('该商品移动/ PC 购买比例')
plt.tight_layout()
filename = '%s_mobile.png' % self.product_id
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.figure(figsize=fig_size)
plt.clf()
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')
def analysis_buy_channel(self):
# self.record_result('用户购买该商品使用的客户端', color = 'black', font_size = 24, strong = True)
obj = self.data_frame['user_client_show']
obj = obj.value_counts()
obj = obj.rename({u'': u'其他，网页端'})
# obj = obj.append(mobile_obj)
# obj.plot(style = 'ro-')
ax = obj.plot(kind='bar', alpha=self.opacity, color=self.color)
ax.set_xticklabels(obj.index, rotation=45 if len(obj.index) > 3 else 0)
# 显示柱状图的百分比
count = obj.sum()
for i, val in enumerate(obj.values):
ax.text(i - 0.25, val, '%.3f%%' % (val * 1.0 / count * 100))
# 尝试将购买渠道和在移动端购买放到一个图中
# plt.subplot(111)
# obj = self.data_frame['is_mobile']
# obj = obj.value_counts()
#
# obj = obj.rename({1: '移动端', 0: 'PC'})
# plt.pie(x = obj.values, autopct = '%.0f%%', radius = 0.3, labels = obj.index)
plt.title('该商品不同客户端的购买数量关系图')
ax.set_xlabel('客户端')
ax.set_ylabel('数量')
filename = '%s_channel.png' % self.product_id
plt.tight_layout()
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.clf()
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')
def analysis_content(self, contents, type):
# 解决京东的评价中包含 &hellip
contents = contents.replace('&hellip', '')
if contents == '':
return
d = '%s/media/mask.png' % settings.BASE_DIR
mask = np.array(Image.open(d))
wordcloud = WordCloud(font_path=self.font_path,
mask=mask).generate(contents)
result = ''
for i, ((word, count), font_size, position, orientation,
color) in enumerate(wordcloud.layout_):
if i <= 3:
result += word + '、'
if type == 'good':
result = '好评数据 关键字：%s' % result
elif type == 'general':
result = '中评数据 关键字：%s' % result
elif type == 'poor':
result = '差评数据 关键字：%s' % result
self.record_result(result, strong=True, color='black', font_size=24)
filename = '%s_%s.png' % (self.product_id, type)
wordcloud.to_image().save('%s/%s' %
(utils.get_save_image_path(), filename))
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')
def analysis_color(self):
# self.record_result('用户购买该商品不同颜色比例', color = 'black', font_size = 24, strong = True)
obj = self.data_frame['product_sku']
logging.warn("obj 111")
logging.warn(obj)
obj = obj.value_counts()
logging.warn("obj 222")
logging.warn(obj)
plt.title('该商品不同颜色购买数量关系图')
ax = plt.subplot(111)
ax.set_xlabel('颜色')
ax.set_ylabel('数量')
obj = obj.rename({'': u'其他'})
logging.warn("obj 333")
logging.warn(obj)
ax = obj.plot(kind='bar', alpha=self.opacity, color=self.color)
# 是否倾斜显示 X 轴标签
if len(ax.containers) > 0:
if len(obj.index) > 5:
xticks_pos = [
1 * patch.get_width() + patch.get_xy()[0]
for patch in ax.containers[0]
]
plt.xticks(xticks_pos, obj.index, rotation=45, ha='right')
else:
xticks_pos = [
0.5 * patch.get_width() + patch.get_xy()[0]
for patch in ax.containers[0]
]
plt.xticks(xticks_pos, obj.index, rotation=0)
# 显示柱状图的百分比
count = obj.sum()
for i, val in enumerate(obj.values):
ax.text(i - 0.25, val, '%.3f%%' % (val * 1.0 / count * 100))
plt.tight_layout()
filename = '%s_color.png' % self.user_id
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.clf()
result = utils.get_image_src(filename=filename)
logging.warn(result)
self.record_result(result, type='image')
def analysis_sell_time(self):
# self.record_result('该商品购买时间、评论时间关系图', color = 'black', font_size = 24, strong = True)
cre_obj = Series(index=self.data_frame['creation_time'], data=1)
cre_obj = cre_obj.resample(rule='M').sum()
cre_obj = cre_obj.fillna(0)
# cre_obj.plot(xticks = cre_obj.index, label = '评论数量', kind = 'line', color = 'orange')
cre_obj.plot(style='o-', xticks=cre_obj.index, label='评论数量')
obj = Series(index=self.data_frame['reference_time'], data=1)
obj = obj.resample(rule='M').sum()
obj = obj.fillna(0)
# obj.plot(xticks = obj.index, label = '购买数量')
ax = obj.plot(style='o-', xticks=obj.index, label='购买数量')
if len(obj.index) <= 5:
ax.set_xticklabels([x.strftime('\n%d\n%m\n%Y') for x in obj.index])
else:
count = len(obj.index)
if count <= 10:
ax.set_xticklabels([
x.strftime('\n%d\n%m\n%Y') if i % 2 == 0 else ''
for i, x in enumerate(obj.index)
])
else:
ax.set_xticklabels([
x.strftime('\n%d\n%m\n%Y')
if i % 4 == 0 or i == (len(obj.index) - 1) else ''
for i, x in enumerate(obj.index)
])
plt.title('该商品购买时间、评论时间关系图')
ax.set_xlabel('时间')
ax.set_ylabel('购买/评论数量')
plt.tight_layout()
plt.legend()
filename = '%s_time.png' % self.product_id
plt.savefig('%s/%s' % (utils.get_save_image_path(), filename))
plt.clf()
result = utils.get_image_src(filename=filename)
self.record_result(result, type='image')