コード例 #1
0
 def test_tuple_label_2(self):
     "io with tuple labels #2"
     label = [(1, datetime.date(2001, 1, 10)),
              (2, datetime.date(2012, 1, 1))]
     d = la.larry([[1, 2], [3, 4]], [label, label])
     io = la.IO(self.filename)
     io['a'] = d
     a = io['a'][:]
     assert_larry_equal(a, d)
コード例 #2
0
ファイル: io_roundtrip.py プロジェクト: zonca/pandas
def larry_roundtrip(filename, lar1, lar2):
    io = la.IO(filename)
    io['lar1'] = lar1
    io['lar2'] = lar2
    lar1 = io['lar1']
    lar2 = io['lar2']
コード例 #3
0
def PortfolioStatsOnDate(directory_name, file_name, params, StatDate):

    ######################
    ### Input parameters
    ######################

    (shortname, extension) = os.path.splitext(file_name)

    #print "file name for symbols = ","_"+shortname+"_"
    #print "file type for symbols = ",extension

    # associate directory_name and file_name with HDF5 file on disk.
    if shortname == "symbols":
        listname = "TAA-Symbols"
    elif shortname == "cmg_symbols":
        listname = "CMG-Symbols"
    elif shortname == "Naz100_Symbols":
        listname = "Naz100-Symbols"
    elif shortname == "biglist":
        listname = "biglist-Symbols"
    elif shortname == "ETF_symbols":
        listname = "ETF-Symbols"
    elif shortname == "ProvidentFundSymbols":
        listname = "ProvidentFund-Symbols"
    elif shortname == "sp500_symbols":
        listname = "SP500-Symbols"
    else:
        listname = shortname

    #hdf5_directory = os.getcwd()+"\\symbols"
    hdf5_directory = os.path.join(os.getcwd(), "symbols")
    hdf5filename = os.path.join(hdf5_directory, listname + "_.hdf5")

    # get current list of symbols
    filename = os.path.join(directory_name, file_name)
    symbols = readSymbolList(filename, verbose=True)

    # loop through symbols to create list of stats
    # - use list (with length same as symbols length) of lists (stats per symbol)
    allSymbolStats = []
    # get 2D quotes from labelled array on disk
    io = la.IO(hdf5filename)
    x = io[listname]
    ##print "x shape = ",x.shape
    # get date labels from labelled array on disk
    date = x.getlabel(1)
    dates = []
    for i in range(len(date)):
        datestr = date[i]
        date_newformat = datetime.date(
            *[int(val) for val in datestr.split('-')])
        dates.append(date_newformat)
    datearray = np.asarray(dates)
    # find index to desired date
    ##print " *******************"
    ##print " ******************* StatDate, and final date in hdf5 file = ", StatDate, date[-1], datearray[-1], type(StatDate), type(date[-1]), type(datearray[-1])
    ##print " *******************"
    dateIndex = x.labelindex(str(StatDate), axis=1)
    ##print " FOUND? index for date & date = ", dateIndex, datearray[dateIndex]
    ##print "last 5 days in datearray = ", datearray[-5:]
    # loop through current symbols and generate stats
    for i, isymbol in enumerate(symbols):
        symbolStats = []
        # find index to current symbol (the smart way) and copy quotes to 'quote' array
        ##print "symbols[isymbol], isymbol = ", symbols[i], isymbol
        hdf5SymbolIndex = x.labelindex(isymbol, axis=0)
        ##print " FOUND? index for symbol & symbol = ", hdf5SymbolIndex, isymbol

        quote = x[hdf5SymbolIndex, :].copyx()

        # use 3 moving averages (MA's) to determine if uptrending on desired date
        SMA1 = params['MA2factor'] * np.mean(
            quote[dateIndex - int(params['MA1']):dateIndex])  #longest MA
        previousSMA2 = np.mean(quote[dateIndex - int(params['MA2']) -
                                     1:dateIndex - 1])  #shortest MA
        SMA2 = np.mean(quote[dateIndex -
                             int(params['MA2']):dateIndex])  #shortest MA
        SMA3 = np.mean(quote[dateIndex - int(params['MA3']):dateIndex])
        if quote[dateIndex] > SMA1 or (quote[dateIndex] > min(SMA2, SMA3)
                                       and SMA2 > previousSMA2):
            uptrend = 1
        else:
            uptrend = 0
        #print " date, symbol, uptrend flag = ", str(StatDate), isymbol, uptrend

        # Only output data for uptrending stocks
        if uptrend == 1:
            symbolStats.append(isymbol)
            symbolStats.append(uptrend)

            # Calculate Gain (Loss) over LongPeriod ending on StatDate and also LongPeriod days prior
            LongPeriod = params['LongPeriod']
            GainLoss = quote[dateIndex] / quote[dateIndex - LongPeriod]
            GainLossPrevious = quote[dateIndex -
                                     LongPeriod] / quote[dateIndex -
                                                         2 * LongPeriod]
            symbolStats.append(GainLoss)
            symbolStats.append(GainLossPrevious)

            ########################################################################
            ### Calculate downside risk measure for weighting stocks.
            ### Use 1./ movingwindow_sharpe_ratio for risk measure.
            ### Modify weights with 1./riskDownside and scale so they sum to 1.0
            ########################################################################

            riskDownside_min = params['riskDownside_min']
            riskDownside_max = params['riskDownside_max']
            gainloss = np.ones((quote.shape[0]), dtype=float)
            gainloss[1:] = quote[1:] / quote[:-1]
            gainloss[isnan(gainloss)] = 1.
            sharpe = ( gmean(gainloss[dateIndex-LongPeriod:dateIndex])**252 -1. )     \
                       / ( np.std(gainloss[dateIndex-LongPeriod:dateIndex])*sqrt(252) )
            riskDownside = 1. / sharpe
            riskDownside = np.clip(riskDownside, riskDownside_min,
                                   riskDownside_max)
            symbolStats.append(riskDownside)

            # add this stock's ranking data to master list
            # - data includes:  symbol, uptrend state, period gain, previous period gain, riskDownside
            allSymbolStats.append(symbolStats)

    # Use collected data to generate ranks
    allSymbolStatsGainLoss = []
    for i in range(len(allSymbolStats)):
        allSymbolStatsGainLoss.append(allSymbolStats[i][2])
    ranksGainLossIndices = np.argsort(allSymbolStatsGainLoss)
    #print "ranksGainLossIndices = ", ranksGainLossIndices

    for i in range(len(ranksGainLossIndices)):
        allSymbolStats[ranksGainLossIndices[i]].append(i)

    allSymbolStatsGainLossPrevious = []
    for i in range(len(allSymbolStats)):
        allSymbolStatsGainLossPrevious.append(allSymbolStats[i][3])
    ranksGainLossIndicesPrevious = np.argsort(allSymbolStatsGainLossPrevious)

    for i in range(len(ranksGainLossIndicesPrevious)):
        allSymbolStats[ranksGainLossIndicesPrevious[i]].append(i)

    ###
    ###
    """
    for i in range(len(allSymbolStats)):
        print "i, allSymbolStats[i] = ", i, allSymbolStats[i]

    from time import sleep
    sleep(30)
    """
    ###
    ###

    # reverse the ranks (make low ranks are biggest gainers)
    # - once finished, remember that lower rank is best performer
    ranksGainLoss = []
    for i in range(len(allSymbolStats)):
        ranksGainLoss.append(allSymbolStats[i][5])
    ranksGainLoss = np.array(ranksGainLoss)
    """
    for i in range(len(ranksGainLoss)):
        print "i, ranksGainLoss[i] = ", i, ranksGainLoss[i]
    """
    maxRank = np.max(
        ranksGainLoss
    )  # use later to reverse ranks so that smallest rank is best performer
    ranksGainLoss -= maxRank - 1
    ranksGainLoss *= -1
    ranksGainLoss += 2

    ranksGainLossPrevious = []
    for i in range(len(allSymbolStats)):
        ranksGainLossPrevious.append(allSymbolStats[i][6])
    ranksGainLossPrevious = np.array(ranksGainLossPrevious)
    maxRank = np.max(
        ranksGainLossPrevious
    )  # use later to reverse ranks so that smallest rank is best performer
    ranksGainLossPrevious -= maxRank - 1
    ranksGainLossPrevious *= -1
    ranksGainLossPrevious += 2

    # weight deltaRank for best and worst performers differently
    rankoffsetchoice = params['numberStocksTraded']
    rankThresholdPct = params['rankThresholdPct']
    delta = -1.0 * (ranksGainLoss - ranksGainLossPrevious) / (ranksGainLoss +
                                                              rankoffsetchoice)
    """
    # if rank is outside acceptable threshold, set deltarank to zero so stock will not be chosen
    #  - remember that low ranks are biggest gainers
    rankThreshold = (1. - rankThresholdPct) * ( ranksGainLoss.max() - ranksGainLoss.min() )
    for ii in range(ranksGainLoss.shape[0]):
        if monthgainloss[ii] > rankThreshold :
            delta[ii,jj] = -monthgainloss.shape[0]/2
    """

    #deltaRank = bn.rankdata(delta,axis=0)
    deltaRank = np.argsort(delta)
    # reverse the ranks (low deltaRank have the fastest improving rank)
    maxrank = np.max(deltaRank)
    deltaRank -= maxrank - 1
    deltaRank *= -1
    deltaRank += 2

    ###
    """
    print ""
    print "delta = ", delta
    print ""
    print "deltaRank = ", deltaRank
    print ""
    for i in range(len(delta)):
        print "i, allSymbolStatsGainLossPrevious, allSymbolStatsGainLoss, ranksGainLossPrevious, ranksGainLoss, delta, deltaRank = ",i, allSymbolStatsGainLossPrevious[i], allSymbolStatsGainLoss[i], ranksGainLossPrevious[i], ranksGainLoss[i], delta[i], deltaRank[i]
    print "shapes of allSymbolStats, delta, deltaRank = ", len(allSymbolStats), len(delta), len(deltaRank)
    """
    ###
    for i in range(len(deltaRank)):
        allSymbolStats[deltaRank[i] - 1].append(i)
    """
    for i in range(len(allSymbolStats)):
        print "i, allSymbolStats[i] = ", i, allSymbolStats[i]
        """

    from time import sleep
    #sleep(30)

    return allSymbolStats