isTr = 1
for i in range (2,3) :
X = read_data("Xtr"+str(i), isTr)
Y = read_data("Ytr"+str(i), isTr)
max_info = ""
max_predic = 0
Y['Bound'][Y['Bound'] == 0] = -1
#f= open("/Users/noch/Documents/workspace/data_challenge/result/console_svm_ker_gaussi_C_big.txt","a+")
#f= open("/home/jibril/Desktop/data_challenge/result/console_svm_ker_gaussi.txt","a+")
print("\n testing on Xtr" +str(i)+ ", Ytr" +str(i))
for k in range(2,5):
print("\n number of char:"+str(k+1))
data_new = prepare_data_div(X, k+1)
data_new['Bound'] = Y['Bound']
data_train, data_test = split_data(data_new, 70)
X_train = data_train.iloc[:,:-1]
Y_tr = pd.DataFrame.as_matrix(data_train['Bound']).astype(float).tolist()
X_te = pd.DataFrame.as_matrix(data_test.iloc[:,:-1])
Y_te = pd.DataFrame.as_matrix(data_test['Bound']).astype(float).tolist()
print("\n finished preparing number of char:" + str(k+1))
gamma_arr = [100, 20, 10, 1, 0.1, 0.01]
nm_char = [6, 6, 5]
lmda = [10**(-5), 0.0001, 10**(-5)]
epoch = [400000, 300000, 300000]
for i in range(3):
isTr = 1
Xtr = read_data("Xtr" + str(i), isTr)
Ytr = read_data("Ytr" + str(i), isTr)
Ytr['Bound'][Ytr['Bound'] == 0] = -1
isTr = 0
Xte = read_data("Xte" + str(i), isTr)
Xte['Id'] = pd.DataFrame({'Id': range(i * 1000, (i + 1) * 1000)})
print("preparing data:" + str(i))
Xtr_p = prepare_data_div(Xtr, nm_char[i])
Xtr_p['Bound'] = Ytr['Bound']
Xte_p = prepare_data_div(pd.DataFrame(Xte['DNA']), nm_char[i])
Xte_p['Id'] = Xte['Id']
Xtr_p = Xtr_p.sample(frac=1)
X_tr = pd.DataFrame.as_matrix(Xtr_p.iloc[:, :-1])
Y_tr = pd.DataFrame.as_matrix(Xtr_p['Bound']).astype(float).tolist()
print("training on data:" + str(i))
w, b = pegasos_(X_tr, Y_tr, lmda[i], epoch[i])
print("testing on data:" + str(i))
result = test_with_id(w, b, Xte_p)
Alpha = []
Y_all_te = []
Y_all_tr = []
D = 0
epoch = [500000, 400000, 400000, 300000, 200000, 500000]
lmda = [1e-05, 1e-05, 0.0001, 0.0001, 0.0001, 1e-05]
num_char = [5, 5, 5, 5, 5, 6]
for t in range(5):
epsl = 0
Z = 0
print("... is training on classifier pegasos" + str(t))
X_tr = prepare_data_div(X_train, num_char[t])
X_tr = pd.DataFrame.as_matrix(X_tr)
D = 1 / len(X_tr)
X_te = prepare_data_div(X_test, num_char[t])
X_te = pd.DataFrame.as_matrix(X_te)
#based-classifier
w, b = pegasos_(X_tr, Y_tr, lmda[t], epoch[t])
Y_pre_tr = predict_pegasos(w, b, X_tr)
Y_pre_te = predict_pegasos(w, b, X_te)
predicted_sco_tr = accuracy_score(Y_pre_tr, Y_tr,
normalize=False) / len(Y_pre_tr)
print("predicted_score_tr:" + str(predicted_sco_tr))