np.savetxt("datasets//predictions//VGG19_" + ver + "_predictions.csv",
pred,
fmt='%1.18f',
delimiter=',')
def round_nearest(x, a):
return np.round(x / a) * a
round_pred = round_nearest(pred, 0.05)
#Check the mse or tau-b score
###Write code here to evaluate the classifier
#The predprob function comes directly from the challenge organizers
pred_prob = predprob(cellularity, pred)
print("Prediction probability score: " + str(pred_prob))
tau_b, p_value = stats.kendalltau(pred[valind], cellularity[valind])
np.savetxt("SPIE_truth_val.csv", cellularity, fmt='%1.18f', delimiter=',')
#Plot
plt.scatter(cellularity[valind], pred[valind])
plt.xlabel("Ground truth")
plt.ylabel("Model prediction")
plt.savefig("datasets//predictions//VGG19_" + ver + "_val_graph.png", dpi=150)
plt.show()
##Make nice results table
#plain = pd.DataFrame()
#plain['slide'] = image_slide
#plain['image'] = image_region
spearArray = []
msqArray = []
predprobArray = []
for j in range(1, 5):
pred1 = np.array(
pd.read_csv(
pathPrefix +
"OneDrive - TU Eindhoven\\Vakken\\2018-2019\\Kwart 4\\BEP\\datasets\\predictions\\"
+ name + "_" + str(j) + "_predictions.csv",
header=None))
pred = []
for i, val in enumerate(pred1):
pred.append(val[0])
pred = np.array(pred)
pred_prob = predprob(cellularity[testind], pred[testind])
print("Prediction probability score for {0}_{2}: {1}".format(
name, pred_prob, j))
tau_b, p_value_tau_b = stats.kendalltau(pred[testind],
cellularity[testind])
spearman_correlation, p_value_spcor = stats.spearmanr(
pred[testind], cellularity[testind])
msq = mean_squared_error(
cellularity[testind],
pred[testind]) #Determine standard deviation as well
print("Tau: {0}\nSpearman Correlation: {1}\nMean Squared Error: {2}".
format(tau_b, spearman_correlation, msq))
tauArray.append(tau_b)
spearArray.append(spearman_correlation)
msqArray.append(msq)