Data_Invaders| Chasers of the Lost Data
Team Updates
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
| import pandas | |
| import keras | |
| from keras.layers import Dense, Dropout, BatchNormalization | |
| import csv,math | |
| import numpy as np | |
| import matplotlib.pyplot as plt | |
| from sklearn import preprocessing | |
| from sklearn.preprocessing import MinMaxScaler | |
| from sklearn.metrics import mean_squared_error | |
| seed = 420 | |
| np.random.seed(seed) | |
| scaler = MinMaxScaler(feature_range=(0, 1)) #define normalization | |
| #model function | |
| def base_model(): #create model | |
| model = keras.Sequential() | |
| model.add(Dense(256,activation='relu',input_shape=(1,4))) | |
| model.add(Dense(128, activation='relu')) | |
| model.add(Dense(64,activation='relu')) | |
| model.add(Dense(32,activation='relu')) | |
| model.add(Dense(16,activation='relu')) | |
| model.add(Dense(8,activation='relu')) | |
| model.add(Dense(4,activation='relu')) | |
| model.add(Dense(1, activation='sigmoid')) | |
| model.compile(loss='mean_squared_error',optimizer='adam') | |
| return model | |
| #data function | |
| def matrix(): #extract the data from excel, and return, all data, train input, train output, test input, test output | |
| results = [] #excel table | |
| with open('input.csv','rt', encoding="ascii") as csvfile: #load excel file | |
| reader = csv.reader(csvfile, delimiter=',', quotechar='|') | |
| for row in reader: | |
| results.append(row) | |
| table=[] #only necessary data | |
| for row in range(1,len(results)): #load only data thats important | |
| if(results[row][12]!=''): | |
| table.append([results[row][4],results[row][5],results[row][7],results[row][8],results[row][12]]) | |
| table=np.float64(table) #convert to numpy matrix | |
| table=np.delete(table,25,axis=0) #delete one zero value | |
| table = scaler.fit_transform(table) #normalize | |
| train_input,train_output = [],[] | |
| precentage_of_train_data = 0.92 #precentage of all data that will be used for training | |
| for row in range(0,int(len(table)*precentage_of_train_data)): #load train data | |
| train_input.append([table[row][0],table[row][1],table[row][2],table[row][3]]) #input | |
| train_output.append([table[row][4]]) #output | |
| test_input, test_output = [], [] | |
| for row in range(int(len(table)*precentage_of_train_data),len(table)): #load test data | |
| test_input.append([table[row][0],table[row][1],table[row][2],table[row][3]]) #input | |
| test_output.append([table[row][4]]) #output | |
| #convert to other shape | |
| train_input=np.expand_dims(train_input,1) | |
| train_output=np.expand_dims(train_output,1) | |
| test_input=np.expand_dims(test_input,1) | |
| test_output=np.expand_dims(test_output,1) | |
| return table, train_input, train_output, test_input, test_output | |
| model = base_model() | |
| print(model.summary()) | |
| table, tx, ty, dx, dy = matrix() | |
| model.fit(tx, ty, validation_data = (dx,dy), epochs=300, batch_size=12, verbose=2) | |
| #define predictions and true data | |
| test_predict = model.predict(dx.copy()) #predictions for testing data | |
| train_predict = model.predict(tx.copy()) #predictions for training data | |
| #table[:,4] - true results | |
| #format predicted data | |
| table_predicted = table.copy() #create new table for predicted data | |
| for i in range(len(train_predict)): #fill up first rows with train data because thats how we splitted them | |
| table_predicted[i,4]=train_predict[i] | |
| for i in range(len(train_predict),len(table)-1): #and second with test data | |
| table_predicted[i,4]=test_predict[i-len(train_predict)] | |
| #inverse transform normalized data | |
| table = scaler.inverse_transform(table.copy()) | |
| table_predicted = scaler.inverse_transform(table_predicted.copy()) | |
| #calculate root mean squared error | |
| score = math.sqrt(mean_squared_error(table_predicted[:,4],table[:,4])) | |
| print('RMSE:') | |
| print(score) | |
| #graph data | |
| x=np.arange(len(table)).astype(float) #define x axis just for row number of result | |
| plt.scatter(x,table[:,4], label = 'True values') #scatter true results | |
| plt.scatter(x,table_predicted[:,4],alpha=0.6, marker=11, label='Predicted values') #scatter predicted results | |
| plt.axis([0,len(table), #x axis | |
| min([min(table[:,4]),min(table_predicted[:,4])])*1.1, #y axis min | |
| max([max(table[:,4]),max(table_predicted[:,4])])*1.1]) #y axis max | |
| plt.axvline(x=len(train_predict)-0.5,color='r',linestyle='--') #draws line to separate train prediction data and test prediction data | |
| plt.text(len(train_predict), max([max(table[:,4]),max(table_predicted[:,4])])/2, | |
| "Test data", rotation=90, verticalalignment='center') | |
| plt.text(len(train_predict)-2, max([max(table[:,4]),max(table_predicted[:,4])])/2, | |
| "Train data", rotation=90, verticalalignment='center') | |
| #axis and legend | |
| plt.xlabel('Result number') | |
| plt.legend(loc='upper left') | |
| plt.ylabel('Result value') | |
| plt.show() | |
Mihajlo Trajković
New poster
Mihajlo Trajković
Working
Mihajlo Trajković
We are on!!
Mihajlo Trajković