public static List <DayEntry> RetrieveEnties(string SearchTerm, string StockCode, DateTime FromDate, DateTime?ToDate = null)
{
var result = new List <DayEntry>();
if (ToDate == null)
{
ToDate = DateTime.Now;
}
using (var conn = new SqlCeConnection(_sqlConnection))
{
var comm = new SqlCeCommand(@"SELECT * FROM Tweets
WHERE Text LIKE @Search
AND CreatedAt >= @FromDate
AND CreatedAt <= @ToDate
AND Positive IS NOT NULL
AND Language = 'English'", conn);
comm.Parameters.AddWithValue("@Search", "%" + SearchTerm + "%");
comm.Parameters.AddWithValue("@FromDate", FromDate.AddDays(-_dayThreshold));
comm.Parameters.AddWithValue("@ToDate", ToDate);
var adpt = new SqlCeDataAdapter(comm);
var data = new DataTable();
adpt.Fill(data);
var currentDate = FromDate;
while (currentDate <= ToDate)
{
var closing = DataRetrieve.GetStockQuote(StockCode, currentDate).Close;
if (closing == 0)
{
currentDate = currentDate.AddDays(1);
continue;
}
var dayData = (from row in data.AsEnumerable()
where row.Field <DateTime>("CreatedAt") > currentDate.AddDays(-_dayThreshold) &&
row.Field <DateTime>("CreatedAt") <= currentDate
select row);
var followerCount = (from row in dayData
select row.Field <int>("UserFollowers") + 1)
.Sum();
var positive = (from row in dayData
select row.Field <double>("Positive"))
.Sum() / dayData.Count();
var maxDensity = 0d;
for (int i = 0; i <= 24 * _dayThreshold; i++)
{
var density = (from row in dayData
where row.Field <DateTime>("CreatedAt") >= currentDate.AddDays(-_dayThreshold).AddHours(i) &&
row.Field <DateTime>("CreatedAt") < currentDate.AddDays(-_dayThreshold).AddHours(i + 1)
select row.Field <double>("Positive"))
.Sum();
maxDensity = Math.Max(maxDensity, density);
}
var entry = new DayEntry
{
Date = currentDate.Date,
ClosingPrice = closing,
FollowerCount = followerCount,
MaxTweetDensity = maxDensity,
PositiveSentiment = positive
};
result.Add(entry);
currentDate = currentDate.AddDays(1);
}
}
return(result);
}
public static double PredictStockPrice(string SearchTerm, string StockCode, DateTime FromDate, DateTime ToDate)
{
Console.WriteLine("Retrieving historic data ({0} days available)", ToDate.Subtract(FromDate).Days);
var samples = RetrieveEnties(SearchTerm, StockCode, FromDate, ToDate);
Console.WriteLine("Calculating multiple regression");
int nAttributes = 3;
int nSamples = samples.Count;
double[,] tsData = new double[nSamples, nAttributes];
double[] resultData = new double[nSamples];
for (int i = 0; i < samples.Count; i++)
{
tsData[i, 0] = samples[i].FollowerCount;
tsData[i, 1] = samples[i].MaxTweetDensity;
tsData[i, 2] = samples[i].PositiveSentiment;
resultData[i] = samples[i].ClosingPrice;
}
double[] weights = null;
int fitResult = 0;
alglib.lsfit.lsfitreport rep = new alglib.lsfit.lsfitreport();
alglib.lsfit.lsfitlinear(resultData, tsData, nSamples, nAttributes, ref fitResult, ref weights, rep);
Console.WriteLine("Retrieving latest tweets");
var tweets = DataRetrieve.SearchTwitter(SearchTerm);
var dayData = from t in tweets
where t.CreatedAt.Date > DateTime.Today.AddDays(-_dayThreshold) &&
t.CreatedAt.Date <= DateTime.Today
select t;
var followerCount = (from row in dayData
select row.FollowersCount + 1)
.Sum();
var positive = (from row in dayData
select row.Positive)
.Sum() / dayData.Count();
var maxDensity = 0d;
var time = DateTime.Now.AddHours(7);
for (int i = 0; i <= 24 * _dayThreshold; i++)
{
var select = (from row in dayData
where row.CreatedAt >= time.AddDays(-_dayThreshold).AddHours(i) &&
row.CreatedAt < time.AddDays(-_dayThreshold).AddHours(i + 1)
select row.Positive);
var density = select.Sum();
maxDensity = Math.Max(maxDensity, density);
}
var predictedPrice = weights[0] * followerCount +
weights[1] * maxDensity +
weights[2] * positive;
return(predictedPrice);
}
