def recommend(user_input, line_bot_api, user_id):
#### User Input ####
extraPortfolio_ratio = user_input.extraPortfolio_ratio
user_selectFunds = user_input.user_selectFunds
user_selectFund_weights = user_input.user_selectFund_weights
user_recommend_num = user_input.user_recommend_num
#### Transform user_selectFund_weights to user_portfolio_weights_arr ####
## Description: Multiply (1-extraPortfolio_ratio) to user_selectFund_weights
user_portfolio_weights_arr = (
1 - extraPortfolio_ratio) * np.array(user_selectFund_weights)
user_portfolio_weights = user_portfolio_weights_arr.tolist()
#### Get the Funds' data ####
## Note: Has removed the funds that were closed or have few data
message = TextSendMessage(text="讀取基金資料\U0010005E...")
line_bot_api.push_message(user_id, message)
df = readcsv.read_df(FUNDS_PREPROC_CSV_NAME)
## Check if user_selectFunds are in df.columns
for fund in user_selectFunds:
if not (fund in df.columns):
return None, None
#### Get the original_cluster_dict information ####
original_cluster_dict = {}
with open('original_cluster_dict' + '.pkl', 'rb') as f:
original_cluster_dict = pickle.load(f)
#### Get the bestFund_dict information ####
bestFund_dict = {}
new_bestFund_dict = {}
with open('bestFund_dict' + '.pkl', 'rb') as f:
bestFund_dict = pickle.load(f)
#### Get the bestFund_dict excluding the groups that user_selectFund was in ####
message = TextSendMessage(text="Pre-processing\U0010005E...")
line_bot_api.push_message(user_id, message)
new_bestFund_dict = calculate.cluster_preproc(original_cluster_dict,
bestFund_dict,
user_selectFunds)
#### Pick the top 5 of bestFunds ####
candidate_list = calculate.pick_candidate(new_bestFund_dict, candidate_num)
#### Calculate the orginal protfolio ####
portfolio_list = []
new_returns = calculate.newFund_return(df, user_selectFunds,
user_selectFund_weights)
orginal_q = calculate.index_q(new_returns, ivt_loop_num)
co_risk = calculate.cal_co_risk(user_selectFunds, user_selectFund_weights,
df, "downside")
co_return = calculate.cal_co_return(user_selectFunds,
user_selectFund_weights, df)
# print(co_risk,co_return)
orginal_protfolio = Portfolio(orginal_q, user_selectFunds,
user_selectFund_weights, co_return / co_risk)
portfolio_list.append(orginal_protfolio)
#### Calculate protfolios with extra funds ####
message = TextSendMessage(text="計算中(這可能會花點時間\U0010005E)...")
line_bot_api.push_message(user_id, message)
for i in range(1, user_recommend_num + 1):
portfolio = calculate.portfolio_alloc(df, candidate_list, i,
user_selectFunds,
user_portfolio_weights,
extraPortfolio_ratio,
ivt_loop_num)
if portfolio != None:
portfolio_list.append(portfolio)
recommend_portfolio = min(portfolio_list, key=attrgetter('index_q'))
number_to_name = {}
with open('label2name_dict' + '.pkl', 'rb') as f:
number_to_name = pickle.load(f)
temp_labels = []
for item in orginal_protfolio.labels:
temp_labels.append(
number_to_name.get(item, "Unkown name") + "(" + item + ")")
orginal_protfolio.labels = temp_labels
# ori = rec
if len(orginal_protfolio.labels) == len(recommend_portfolio.labels):
return orginal_protfolio, recommend_portfolio
temp_labels = []
for item in recommend_portfolio.labels:
temp_labels.append(
number_to_name.get(item, "Unkown name") + "(" + item + ")")
recommend_portfolio.labels = temp_labels
return orginal_protfolio, recommend_portfolio
