import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
import xgboost
import pandas as pd
import shap

# train a "dummy" XGBoost model to get initial SHAP values
X, y = shap.datasets.boston()
model = xgboost.XGBRegressor().fit(X, y)
explainer = shap.Explainer(model)
shap_values = explainer(X)

# get LIME values of experiment
data = pd.read_csv("Data/transfer_complete_.csv")
data_subset = data[['trustee', 'Competitiveness_Score', 'Openness', 'Conscentiousness',
       'Agreeableness', 'Neuroticism', 'Extraversion', 'gender_y',
       'financed_by_parents', 'younger_siblings', 'older_siblings',
       'Competitiveness_Score_importance', 'Openness_importance',
       'Conscentiousness_importance', 'Agreeableness_importance',
       'Neuroticism_importance', 'Extraversion_importance',
       'gender_importance', 'patience', 'younger_siblings_importance',
       'older_siblings_importance']]
data_subset = data_subset.drop_duplicates()
data_subset = data_subset.set_index('trustee')
data_characteristics = data_subset.loc[:, 'Competitiveness_Score':'older_siblings']
data_char_importance = data_subset.loc[:, 'Competitiveness_Score_importance':'older_siblings_importance']

# get mean absolute LIME values for each characteristic
mean_abs_LIME_val = []
for (columnName, columnData) in data_char_importance.iteritems():
    absolute_LIME_val = [abs(l) for l in columnData]
    val = str(round(sum(absolute_LIME_val)/len(absolute_LIME_val), 2))
    mean_abs_LIME_val.append(" ("+val+")")

# overwrite SHAP values
# ...base_values are 0
for i in range(0, len(shap_values)):
    shap_values.base_values[i] = 1
# ...LIME feature importance
shap_values.values = data_char_importance.values
# ...borrower characteristics (values)
shap_values.data = data_characteristics.values
# ...characteristics (names)
shap_values.feature_names = ['Competitiveness' + mean_abs_LIME_val[0], 'Openness' + mean_abs_LIME_val[1],
                             'Conscentiousness' + mean_abs_LIME_val[2], 'Agreeableness' + mean_abs_LIME_val[3],
                             'Neuroticism' + mean_abs_LIME_val[4], 'Extraversion' + mean_abs_LIME_val[5],
                             'Gender' + mean_abs_LIME_val[6], 'Patience' + mean_abs_LIME_val[7],
                             'Younger siblings' + mean_abs_LIME_val[8], 'Older siblings' + mean_abs_LIME_val[9]]


# plot "shap" aka LIME values
plt.style.use('science')
plt.figure()
shap.summary_plot(shap_values, plot_size=(12, 12), show=False, mz_custom = True)

# set plot properties
fig, ax = plt.gcf(), plt.gca()

# set label sizes
for item in ([ax.title, ax.xaxis.label, ax.yaxis.label] +  ax.get_xticklabels() + ax.get_yticklabels()):
    if item == ax.get_xticklabels():
       item.set_fontsize(17)
       continue
    item.set_fontsize(20)

# show plot
fig.tight_layout()

# # save to pdf
pp = PdfPages('multipage.pdf')
fig.savefig(pp, format='pdf')
pp.close()
# fig.show()