def deleteEarlyModel(self):
collection = getConnection('mongo')[self.modelColName]
try:
buildids = []
header = self.HEADERS_JSON
for item in collection.find().sort("time", ASCENDING):
buildids.append(item)
# print(buildids)
if len(buildids) > 0:
buildurl = self.endpoint + '/models/%s/builds/%s' % (self.modelid, buildids[0]["buildId"])
resp = requests.delete(buildurl, headers=header)
print("delete status = ", resp.status_code)
# result = json.loads(resp.text)
if resp.status_code == 200 or resp.status_code == 204:
collection.delete_one(buildids[0])
except:
print_exc()
def generateUsageFile(self, size, users = [], filename = 'usage.txt'):
collection = getConnection('mongo')[self.colName]
# temp = tempfile.TemporaryFile()
try:
items = []
for news in collection.find({}):
items.append(news)
with open(filename, 'wb') as file:
for i in range(0, size): # TODO: look up usage file format
file.write((','.join([
users[random.randint(0, len(users)-1)].replace('@','-').replace('.','_'), # userid
str(items[random.randint(0, len(items)-1)]["_id"]), # itemid
datetime.now().strftime(self.timeFormat),
"Purchase"
])+"\n").encode())
except:
print_exc()
finally:
# temp.close()
pass
def buildRelated(self):
collection = getConnection('mongo')[self.colName]
model = gensim.models.Word2Vec.load(self.word2vecModel)
try:
for news in collection.find({ "related_words": { "$exists": 0 } }, projection={"keywords":1, "title": 1, "_id": 1}):
relatedWords = set()
for keyword in news["keywords"]:
try:
relatedWords |= set(map(lambda t: t[0], model.most_similar(keyword)))
except KeyError:
print("KeyError:", keyword)
continue
print(news["title"])
collection.update_one({"_id": news["_id"]},
{"$set": { "related_words": list(relatedWords) }})
except:
print_exc()
finally:
model = None
gc.collect()
def triggerBuild(self):
collection = getConnection('mongo')[self.modelColName]
header = self.HEADERS_JSON
buildurl = self.endpoint + '/models/%s/builds?' % (self.modelid)
body = json.dumps({
"description": "Simple recomendations build",
"buildType": "recommendation",
"buildParameters": {
"recommendation": {
"numberOfModelIterations": 40,
"numberOfModelDimensions": 20,
"itemCutOffLowerBound": 1,
"itemCutOffUpperBound": 10,
"userCutOffLowerBound": 0,
"userCutOffUpperBound": 0,
"enableModelingInsights": False,
"useFeaturesInModel": True,
"modelingFeatureList": "tag",
"allowColdItemPlacement": True,
"enableFeatureCorrelation": True,
"reasoningFeatureList": "tag",
"enableU2I": True
}
}
})
try:
resp = requests.post(buildurl, body, headers=header)
result = json.loads(resp.text)
print("url = ", buildurl)
print(result)
if resp.status_code == 202: # success
collection.insert({
"buildId": result["buildId"],
"time": datetime.now().strftime(self.timeFormat),
"modelId": self.modelid,
"token": self.token
})
except:
print_exc()
def updateCatalog(self):
catalogurl = self.endpoint + '/models/%s/catalog' % self.modelid
header = self.HEADERS_STREAM
collection = getConnection('mongo')[self.colName]
temp = tempfile.TemporaryFile()
try:
for news in collection.find({ "uploaded": False }):
tags = []
itemName = "||".join([news["genre"], news["title"], news["imgurls"][0] if len(news["imgurls"]) > 0 else ""])
for tag in news["keywords"]:
tags.append("tag="+tag)
temp.write((','.join([str(news["_id"]), itemName, news["genre"], ""] + tags)+"\n").encode())
temp.seek(0)
resp = requests.patch(catalogurl, data=temp, headers=header)
result = json.loads(resp.text)
print("result = " + str(result))
if resp.status_code == 200:
collection.update_many({ "uploaded": False },
{ "$set": { "uploaded": True } }, upsert=False)
except:
print_exc()
finally:
temp.close()