Spaces:
Running
Running
| import sys | |
| sys.path.append('../XGBoost_Prediction_Model/') | |
| import warnings | |
| warnings.filterwarnings("ignore") | |
| import Predict | |
| import torch | |
| import numpy as np | |
| import os | |
| from os.path import isfile, isdir, join | |
| from Magazine_Optimization import * | |
| import time | |
| mypath = '../XGBoost_Prediction_Model/Magazine_Optimization_Demo/Magazines' | |
| results = {} | |
| for f in os.listdir(mypath): | |
| if isdir(join(mypath, f)): | |
| print('Currently processing Magazine '+f+'......') | |
| path_temp = join(mypath, f) | |
| dir_list = [] | |
| for sub_f in os.listdir(path_temp): | |
| if isdir(join(path_temp, sub_f)): | |
| sub_path_temp = join(path_temp, sub_f) | |
| if (sub_f.split('_')[0]) == 'Jpg': | |
| dir_list = os.listdir(sub_path_temp) | |
| dir_list.sort() | |
| for i in range(len(dir_list)): | |
| dir_list[i] = join(sub_path_temp,dir_list[i]) | |
| else: | |
| Slots = torch.load(join(sub_path_temp,'Slots')).astype('int32') | |
| Sizes = torch.load(join(sub_path_temp,'surfaces')) | |
| Product_Groups = torch.load(join(sub_path_temp,'Prod_Cat')) | |
| Textboxes = torch.load(join(sub_path_temp,'Textboxes')) | |
| Obj_and_Topics = torch.load(join(sub_path_temp,'Obj_and_Topics')) | |
| Ad_embeddings = torch.load(join(sub_path_temp,'Ad_Emb')) | |
| Ctpg_embeddings = torch.load(join(sub_path_temp,'Ctpg_Emb')) | |
| start = time.time() | |
| Costs = np.ones(len(dir_list)) #np.arange(len(dir_list),0,-1) | |
| Ad_Gaze, Brand_Gaze, Double_Page_Ad_Attention, Double_Page_Brand_Attention, Assign_ids = Preference_Matrix(dir_list, Slots, Product_Groups, Sizes, | |
| Ad_embeddings=Ad_embeddings, Ctpg_embeddings=Ctpg_embeddings, | |
| Textboxes=Textboxes, Obj_and_Topics=Obj_and_Topics, | |
| Costs=Costs, Method='XGBoost') | |
| #Assignement Problem | |
| workers = [] | |
| jobs = [] | |
| N = (Ad_Gaze.shape)[0] | |
| for i in range(N): | |
| workers.append(i+1) | |
| jobs.append(i+1) | |
| max_ad_attention = np.max(Ad_Gaze) | |
| max_brand_attention = np.max(Brand_Gaze) | |
| Ad_Gaze_cost = max_ad_attention - Ad_Gaze | |
| Brand_Gaze_cost = max_brand_attention - Brand_Gaze | |
| Prob_solved_Ad = Assignment_Problem(Ad_Gaze_cost, workers, jobs) | |
| Prob_solved_Brand = Assignment_Problem(Brand_Gaze_cost, workers, jobs) | |
| end = time.time() | |
| assigning_time = end-start | |
| # Print the variables optimized value | |
| print('If based on maximizing Overall Ad Attention: ') | |
| strategy_AG = '' | |
| BG_under_AG_assignment = 0 | |
| for v in Prob_solved_Ad.variables(): | |
| if v.varValue == 1: | |
| curr = (v.name).split('_') | |
| temp = curr[0]+' Ad '+str(Assign_ids[int(curr[1])-1])+' to Counterpage '+str(Assign_ids[int(curr[2])-1]) | |
| BG_under_AG_assignment += Brand_Gaze_cost[int(curr[1])-1,int(curr[2])-1] | |
| strategy_AG += temp+'; ' | |
| print(temp) | |
| # The optimised objective function value is printed to the screen | |
| m_ad = N*max_ad_attention - value(Prob_solved_Ad.objective) + sum(Double_Page_Ad_Attention) | |
| print("Maximized Ad Attention = ", m_ad, " sec.") | |
| print("Maximized Average Ad attention on each Ad = ", (N*max_ad_attention - value(Prob_solved_Ad.objective) + sum(Double_Page_Ad_Attention))/(N + len(Double_Page_Ad_Attention)), " sec.") | |
| print("Original Ad Attention = ", np.trace(Ad_Gaze), " sec.") | |
| print() | |
| # Print the variables optimized value | |
| print('If based on maximizing Overall Brand Attention: ') | |
| strategy_BG = '' | |
| for v in Prob_solved_Brand.variables(): | |
| if v.varValue == 1: | |
| curr = (v.name).split('_') | |
| temp = curr[0]+' Ad '+str(Assign_ids[int(curr[1])-1])+' to Counterpage '+str(Assign_ids[int(curr[2])-1]) | |
| strategy_BG += temp+'; ' | |
| print(temp) | |
| # The optimised objective function value is printed to the screen | |
| m_brand = N*max_brand_attention - value(Prob_solved_Brand.objective) + sum(Double_Page_Brand_Attention) | |
| BG_under_AG_assignment = N*max_brand_attention - BG_under_AG_assignment + sum(Double_Page_Brand_Attention) | |
| print("Maximized Brand Attention = ", m_brand, " sec.") | |
| print("New Brand Gaze under AG assignment = ", BG_under_AG_assignment, " sec.") | |
| print("Maximized Average Brand attention on each Ad = ", (N*max_brand_attention - value(Prob_solved_Brand.objective) + sum(Double_Page_Brand_Attention))/(N + len(Double_Page_Brand_Attention)), " sec.") | |
| print("Original Brand Attention = ", np.trace(Brand_Gaze), " sec.") | |
| print('End of Magazine '+f+'......') | |
| results[f] = {'AG':[strategy_AG,m_ad,np.trace(Ad_Gaze)], 'BG':[strategy_BG,m_brand,np.trace(Brand_Gaze),BG_under_AG_assignment], 'Time':assigning_time} | |
| print() | |
| print() | |
| print() | |
| torch.save(results, '../XGBoost_Prediction_Model/Magazine_Optimization_Demo/results') | |
| # torch.save(results, '../XGBoost_Prediction_Model/Magazine_Optimization_Demo/results_CNN') | |
| print('Summary: ') | |
| for f in list(results.keys()): | |
| print('Magazine '+f+': ') | |
| dict_curr = results[f] | |
| print('Total Time used: ',dict_curr['Time']) | |
| print('Ad Gaze: ') | |
| print('Strategy: '+dict_curr['AG'][0]) | |
| print('max Attention: ',dict_curr['AG'][1]) | |
| print('original Attention: ',dict_curr['AG'][2]) | |
| print('Improvement: ', (dict_curr['AG'][1]-dict_curr['AG'][2])/dict_curr['AG'][2]*100) | |
| print('------------------------') | |
| print('Brand Gaze: ') | |
| print('Strategy: '+dict_curr['BG'][0]) | |
| print('max Attention: ',dict_curr['BG'][1]) | |
| print('original Attention: ',dict_curr['BG'][2]) | |
| print('Attention under AG Assignment: ',dict_curr['BG'][3]) | |
| print('Improvement: ', (dict_curr['BG'][3]-dict_curr['BG'][2])/dict_curr['BG'][2]*100) | |
| print('------------------------') | |
| print() |