

In [1]:

    
%matplotlib inline
import pandas as pd
import sys
import shutil
from alpenglow.experiments import BatchFactorExperiment, ExternalModelExperiment
from alpenglow.evaluation import DcgScore
import numpy as np
import matplotlib
import matplotlib.pyplot as plt



In [2]:

    
data = pd.read_csv('http://info.ilab.sztaki.hu/~fbobee/alpenglow/tutorial_dataset.csv', header=None, names=['time', 'user', 'item'])

alpenglow model



In [3]:

    
exp = ExternalModelExperiment(
    period_length=60 * 60 * 24 * 7 * 4,
    in_name_base="batches/batch",
    mode="read",
)



In [4]:

    
res = exp.run(data)









    



running experiment...



In [5]:

    
res['dcg'] = DcgScore(res)



In [6]:

    
res['dcg'].mean()









    Out[6]:





0.014293958938788652

lightfm



In [7]:

    
exp = ExternalModelExperiment(
    period_length=60 * 60 * 24 * 7 * 4,
    in_name_base="../lightfm/batches/batch",
    mode="read",
)



In [8]:

    
res3 = exp.run(data)









    



running experiment...



In [9]:

    
res3['dcg'] = DcgScore(res3)



In [10]:

    
res3['dcg'].mean()









    Out[10]:





0.01757739322309966

turicreate



In [11]:

    
exp = ExternalModelExperiment(
    period_length=60 * 60 * 24 * 7 * 4,
    in_name_base="../turicreate/batches/batch",
    mode="read",
)
res4 = exp.run(data)
res4['dcg'] = DcgScore(res4)
res4['dcg'].mean()









    



running experiment...






    Out[11]:





0.023746456208509535

libfm



In [12]:

    
exp = ExternalModelExperiment(
    period_length=60 * 60 * 24 * 7 * 4,
    in_name_base="../libfm/batches/batch",
    mode="read",
)
res5 = exp.run(data)
res5['dcg'] = DcgScore(res5)
res5['dcg'].mean()









    



running experiment...






    Out[12]:





0.01446016560516256

alpenglow batch factor model



In [13]:

    
exp = BatchFactorExperiment(
    period_length=60 * 60 * 24 * 7 * 4,
    negative_rate=5,
    learning_rate = 0.05,
    clear_model=True
)
res2 = exp.run(data)









    



running experiment...



In [14]:

    
res2['dcg'] = DcgScore(res2)



In [15]:

    
res2['dcg'].mean()









    Out[15]:





0.013916101948001271

plotting



In [16]:

    
plt.figure(figsize=(15,8))
res.groupby(res.time//(60*60*24)).dcg.mean().plot()
res2.groupby(res2.time//(60*60*24)).dcg.mean().plot()
res3.groupby(res2.time//(60*60*24)).dcg.mean().plot()
res4.groupby(res2.time//(60*60*24)).dcg.mean().plot()
res5.groupby(res2.time//(60*60*24)).dcg.mean().plot()
plt.ylabel('daily average ndcg')
plt.legend(['alpenglow external', 'alpenglow', 'lightfm', 'turicreate', 'libfm'])









    Out[16]:





<matplotlib.legend.Legend at 0x7f23b7c5bda0>



In [ ]: