int main(){
vector<int> percentage;
vector<int> attempts;
vector<int> incorrect;
for(int percent_held=40; percent_held<51; percent_held+=10)
{
percentage.push_back(percent_held);
int num_wrong = 0;
int iterations =0;
int iter = 0;
for(int iter=0; iter<10; iter++)
{
Classifier<NgramVector> classifier;
vector<string> file_names;
file_names.push_back("data/bar-tweets.txt");
file_names.push_back("data/ht-tweets.txt");
ClassifiedDataSet<NgramVector> test;
for(int j=0; j<file_names.size(); j++)
{
//test on bar-tweets.txt
ifstream fin(file_names[j]);
//ifstream fin("data/ht-tweets.txt");
ClassifiedDataSet<NgramVector> v;
map<string, NgramVector> vect_map;
int i = 1;
while(!fin.eof()){
//NgramVector temp(to_string(i++),"2","multi",3);
//get line
string line;
getline(fin, line);
//get tweet_id
string tweet_id = line.substr(0,line.find('\t'));
line.erase(0, tweet_id.size()+1);
//get user_id
string user_id = line.substr(0,line.find('\t'));
line.erase(0, tweet_id.size()+1);
//get tweet
string tweet = line.substr(0,line.find('\t'));
//if user_id not in map
if(vect_map.find(user_id) == vect_map.end())
{
NgramVector temp(tweet_id,user_id,file_names[j],3);
vect_map[user_id] = temp;
//cout << user_id << " " << tweet << endl;// tweet << endl;
}
vect_map[user_id].input_string(tweet);
}
srand (time(NULL));
//create c++ vector of random integers from 0 to vect_map.size()
//want random 20 percent of vectors not included in training set.
unordered_map<int,int> test_set;
for(int i=0; i<= ((double)percent_held/(double)100)*vect_map.size(); i++)
{
int iSecret;
bool found = false;
while(!found)
{
//cout << "a" << endl;
//find random int not yet in test set;
/* initialize random seed: */
/* generate secret number between 1 and 10: */
iSecret = rand() % (vect_map.size()-1);
if(test_set.find(iSecret) == test_set.end())
{
found = true;
test_set[iSecret] = 0;
}
}
//cout << iSecret << endl;
}
int m=0;
for(auto it=vect_map.begin(); it!=vect_map.end(); it++)
{
if(test_set.find(m) == test_set.end())
{
v.push_back(it->second);
//cout << it->second.get_lang() << endl;
}
else
{
test.push_back(it->second);
//cout << "-"<< it->second.get_lang() << endl;
}
m++;
}
//v.info_gain_for_each_value();
//v.print_ordered();
classifier.add_classified_dataset(v, file_names[j]);
}
//classifier.print_ordered();
classifier.calculate_info_gain();
for(int i=0; i<test.size(); i++)
{
iterations++;
cout << "------------------------------" << endl;
string expected = test[i].get_lang();
cout << "exptected: " << expected << endl;
string return_val = classifier.naive_classify(test[i]);
cout << "actual: " << return_val << endl;
if(expected != return_val)
{
cout << "incorrectly guessed value" << endl;
num_wrong++;
}
}
cout << "number wrong for " << iterations << " iterations = " << num_wrong << endl;
//classifier.print_universal_info_gain();
//add error and iterations
}
attempts.push_back(iterations);
incorrect.push_back(num_wrong);
for(int num=0; num<percentage.size(); num++)
{
cout << "==============================================" << endl;
cout << "-----------------------------------" << endl;
cout << "trial percentage: " << percentage[num] << endl;
cout << "attempst : " << attempts[num] << endl;
cout << "incorrect : " << incorrect[num] << endl;
cout << "error : " << incorrect[num]/attempts[num] << endl;
}
}
}
