/// <summary>
/// 将样本集划分为两个子集
/// </summary>
/// <param name="ratioOfFirstGroup">第一个子集的占比</param>
/// <param name="ratioOfSecondGroup">第二个子集的占比</param>
/// <param name="method">切分方法</param>
/// <param name="firstGroup">第一个子集</param>
/// <param name="secondGroup">第二个子集</param>
public void DivideIntoTwo(int ratioOfFirstGroup, int ratioOfSecondGroup, CollectionDividingMethod method, out SampleCollection firstGroup, out SampleCollection secondGroup)
{
this.DivideByRatioAndMethod(ratioOfFirstGroup, ratioOfSecondGroup, method, out List <Sample> samples1, out List <Sample> samples2);
firstGroup = new SampleCollection(samples1);
secondGroup = new SampleCollection(samples2);
}
/// <summary>
/// 算术平均值过滤。窗口长度内取平均值
/// </summary>
/// <param name="windowLengths">各维度的窗口长度</param>
/// <returns>新的样本集</returns>
public SampleCollection MeanFilter(int[] windowLengths)
{
windowLengths.ShouldEachSatisfy((m) => m.ShouldLargerThan(0) && m.ShouldNotLargerThan(this.Count()));
windowLengths.Length.ShouldEquals(this.TotalDimensionCount);
SampleCollection result = this.Clone();
CollectionHelper.ForEachThree(
this.Columns,
result.Columns,
windowLengths,
(srcCol, resultCol, winlength) => {
LimitingAmplifyFilter(srcCol, resultCol, winlength);
});
return(result);
}
/// <summary>
/// 限幅过滤。放弃掉波动过大的样本，用前一个样本代替。
/// </summary>
/// <param name="diffLimitations">波动最大值绝对值的向量. 都必须大于0</param>
/// <returns>新的样本集</returns>
public SampleCollection LimitingAmplifyFilter(IEnumerable <double> diffLimitations)
{
diffLimitations.Count().ShouldEquals(this.TotalDimensionCount);
diffLimitations.ShouldEachSatisfy((item) => item > 0);
SampleCollection result = this.Clone();
CollectionHelper.ForEachThree(
this.Columns,
result.Columns,
diffLimitations,
(srcCol, resultCol, diff) => {
LimitingAmplifyFilter(srcCol, resultCol, diff);
});
return(result);
}
/// <summary>
/// 检查字符串是否是样本集中的字段名
/// </summary>
/// <param name="fieldName"></param>
/// <returns>是字段名返回真，反之假</returns>
public static SampleCollection LoadFromCSVFile(StreamReader reader, int outputDimension = 0, bool ignoreInvalidLine = false, bool firstLineFieldName = true)
{
SampleCollection sc = new SampleCollection();
int lineNum = 0;
string[] fieldNames = null;
bool validLine;
int totalDimension = -1;
foreach (var line in reader.ReadLines())
{
lineNum++;
validLine = true;
var strValues = line.Replace(" ", "").Split(',');
if (totalDimension == -1)
{
totalDimension = strValues.Length;
}
else
{
if (strValues.Length != totalDimension)
{
if (ignoreInvalidLine)
{
validLine = false;
continue;
}
else
{
throw new DimensionNotMatchException("Dimension not match in line:" + lineNum.ToString());
}
}
}
if (lineNum == 1 && firstLineFieldName)
{
fieldNames = strValues;
continue;
}
var values = new double[totalDimension];
for (int i = 0; i < strValues.Length; i++)
{
if (!double.TryParse(strValues[i], out double value))
{
if (ignoreInvalidLine)
{
validLine = false;
break;
}
else
{
throw new Exception("Has non-double value in line:" + lineNum.ToString());
}
}
values[i] = value;
}
if (validLine)
{
sc.Add(new Sample(values, outputDimension, fieldNames));
}
}
return(sc);
}
