ax.tick_params(axis='both', which='major', labelsize=25)
ax.tick_params(axis='both', which='minor', labelsize=25)
ax.set_title('Inverse Participation Ratio week ' + str(w+1), fontsize = 30)
ax.grid()
# graph[countGraph].axhline(y=0, color='k')
# graph[countGraph].axvline(x=0, color='k')
eigenValueList = pca_result[9].explained_variance_
eigenVectorList = pca_result[9].components_
IPRlist = libRMT.IPRarray(eigenValueList,eigenVectorList)
ax.axhline(y=IPRlist['IPR'].mean(), color='k', label='mean value of IPR')
ax.plot(IPRlist['eigenvalue'], IPRlist['IPR'], '-', color ='blue', label='IPR')
ax.legend(loc='upper right')
plt.show()
#outbounce select
a = libRMT.selectOutboundComponents(pcaDataWeeks[9],pca_result[9].explained_variance_)
#eigenvalues
fig = plt.figure(figsize=(40,30),dpi=240)
graph = []
countGraph = 0
num_bins = 100
for w in range(0,12):
ax = fig.add_subplot(3,4,w+1)
graph.append(ax)
graph[countGraph].set_xlabel('eigenvalue λ', fontsize = 15)
graph[countGraph].set_ylabel('P(λ)', fontsize = 15)
graph[countGraph].set_title('Week ' + str(w+1), fontsize = 20)
graph[countGraph].grid()
# graph[countGraph].axhline(y=0, color='k')
# graph[countGraph].axvline(x=0, color='k')
# tempData = tempData.merge(prediction_transition[w+1]['data']['successPassedRate'], left_on = tempData.index, right_on=prediction_transition[w+1]['data']['successPassedRate'].index).set_index('key_0')
temp = FCAMiner.PCAcohortToValue(tempData)
temp1 = temp[1]
pcaResult = temp[0]
# temp1 = temp1.merge(prediction_transition[w+1]['data']['result_exam_1'], left_on = temp1.index, right_on=prediction_transition[w+1]['data']['result_exam_1'].index).set_index('key_0')
pcaDataWeeks.append(temp1)
pca_result.append(pcaResult)
columnsReturn2.append(temp[2])
import libRMT
transitionDataMatrixWeeks_directFollow_standardised_outbound = []
for week in range(0, 12):
transitionDataMatrixWeeks_directFollow_standardised_outbound.append(
libRMT.selectOutboundComponents(pcaDataWeeks[week],
pca_result[week].explained_variance_,
mode="upper"))
transitionDataMatrixWeeks_directFollow_standardised_cleaned = []
for week in range(0, 12):
outBoundComponents = transitionDataMatrixWeeks_directFollow_standardised_outbound[
week].columns
componentToClean = ['pc1']
for c in pcaDataWeeks[week].columns:
if c not in outBoundComponents:
componentToClean.append(c)
transitionDataMatrixWeeks_directFollow_standardised_cleaned.append(
libRMT.cleanEigenvectorEffect(
transitionDataMatrixWeeks_directFollow_standardised[week],
pcaDataWeeks[week], componentToClean, pca_result[week].components_,
0.8, 0.5))
'correct'] / practiceResult.groupby(
[pd.Grouper(key='user'),
pd.Grouper(key='task')]).count()['correct']
cummulativeResult = practiceResultSum.reset_index().groupby(
[pd.Grouper(key='user')]).sum()
# cummulativeResult = practiceResultSum.groupby([pd.Grouper(key='user')]).sum()
cummulativeResult['cumm_practice'] = cummulativeResult[
'correct'] / practiceResult.groupby([pd.Grouper(key='user')
]).count()['date']
cummulativeResult['successPassedRate'] = cummulativeResult['passed'] / (
cummulativeResult['passed'] + cummulativeResult['failed'])
pcaData1 = transitionDataMatrixWeeks[week] #original data - scenario 1
pcaData2 = libRMT.selectOutboundComponents(
pcaDataWeeks[week], eigenValueList[week]
) #filtered pca data to original data - scenario 2 #testing with eigenvalue > 1
pcaData3 = pcaDataWeeks[week] #full pca data - scenario 3
pcaData4 = transitionDataMatrixWeeks_normalised_cleaned[
week] #clean data - scenario 4
pcaData5 = pcaDataWeeks_cleanedData[week] #pca clean data - scenario 5
pcaData6 = libRMT.selectOutboundComponents(
pcaDataWeeks_cleanedData[week], eigenValueList_cleanedData[week]
) #pca filtered clean data - scenario 6
# print(pcaData.columns)
mode = 'transition'
tic = time.time()
test1 = PredictionResult.predict_proba_all_algorithms_data_ready(
pcaData1, excellent, weak, cummulativeResult, mode)