def __init__(self, path_root=''):
self.path_root = path_root
self.sqlite_data = sqlite_data_handler(self.path_root)
self.competition = self.sqlite_data.table_Competitions
self.team = self.sqlite_data.table_Teams
self.submission = self.sqlite_data.table_Submissions
self.competitionhostsegment = self.sqlite_data.table_CompetitionHostSegments
plt.scatter(points['x'], points['y'], linewidths=0.5, s=10)
plt.show()
def describe(self):
for k in np.unique(self.labels):
cluster = self.data[(self.data['cluster'] == k)]
print("Cluster: ", k)
print("Median:")
print(cluster.median())
print("Mean:")
print(cluster.mean())
print("\n")
if __name__ == '__main__':
data = sqlite_data_handler("./data")
processor = pp.preProcessor(data)
users = processor.get_users()
# Select users and features to cluster upon
users = users.loc[
(users['HighestRanking'] < 200) & (users['NumSubmissions'] > 1),
['NumSubmissions', 'NumPosts', 'DaysTillActivity', 'HighestRanking']]
# Select columns to apply log to
users = pu.log_data(users, ['NumSubmissions', 'NumPosts'])
print(users.describe())
users = pu.normalize_data(users)
# Select User features
users = users.sample(frac=0.1)
comps = points.CompetitionId.sort_values().unique()
m = pd.DataFrame(np.zeros((len(users), len(comps))))
m.columns = comps
m.index = users
def to_matrix(row):
m.loc[row.UserId,
row.CompetitionId] = row.Points + sum(m.loc[row.UserId, ])
points = points.sort_values(['CompetitionId', 'UserId'])
points.apply(to_matrix, axis=1)
return m
if __name__ == '__main__':
db = sqlite_data_handler('data')
print("Computing Points...")
points = computePoints(db)
comps = points.groupby(['CompetitionId']).agg({
'Points':
'sum',
'UserId':
'nunique',
'UserRankMultiplier':
'median'
})
# TODO: User Comp matrix with aggregated User Counts
# TODO: 2. aggregated user counts have to be normalized? Mean better?
m = user_comp_matrix(points)
print(m)