public override void Fit(BaseDataSet dataSet)
{
Profiler.BeginSample("SlopeOne.Fit");
var nItems = dataSet.itemCount;
base.Fit(dataSet);
m_Freq = new int[nItems, nItems];
m_Dev = new double[nItems, nItems];
foreach (var(_, uRatings) in dataSet.userRatings)
{
foreach (var(_, i, rUi) in uRatings)
{
foreach (var(_, j, rUj) in uRatings)
{
m_Freq[i, j] += 1;
m_Dev[i, j] += rUi - rUj;
}
}
}
for (var i = 0; i < nItems; i++)
{
m_Dev[i, i] = 0;
for (var j = i + 1; j < nItems; j++)
{
m_Dev[i, j] /= m_Freq[i, j];
m_Dev[j, i] = -m_Dev[i, j];
}
}
m_UserMean = dataSet.allUsers
.Select(u => dataSet.UserRatings(u).Select(rt => rt.rating).Average()).ToArray();
Profiler.EndSample();
}
public CollaborativeFilteringRecommender(BaseAlgorithm algorithm)
{
m_Algorithm = algorithm;
m_DataSet = new ArrayBackedSinglePlayerDataSet(MonsterMatchArrayData.Data,
MonsterMatchArrayData.UserCount, MonsterMatchArrayData.ItemCount);
m_PlayerUserId = MonsterMatchArrayData.PlayerUserId;
Debug.Log($"DefaultRecommender: m_PlayerUserId={m_PlayerUserId}");
m_ApplicationItemIdsToInternalIds = Enumerable.Range(0, MonsterMatchArrayData.ItemCount)
.ToDictionary(i => MonsterMatchArrayData.ForProfiles[i], i => i);
m_InternalIdsToApplicationItemIds = MonsterMatchArrayData.ForProfiles;
}
public virtual void Fit(BaseDataSet dataSet)
{
m_DataSet = dataSet;
}
