predict.py 文件源码

def fmt_numerical(self, data):
        """
        There are two categories of data Porn Sieve gathers:
          1.Tag data, represented as a binary, mostly zero array of numbers
          2. Data which is continuous, such as duration, average review, etc.

        For the tags, I can just use CountVectorizer out-of-the-box, but for
        the other data, we need to put it all together in a list on our own.
        """
        nums = []
        # sorted to ensure the data is always in the same order
        for k in sorted(data.keys()):
            if k in ["feedback", "img"]:
                pass

            elif type(data[k]) == list:
                pass

            elif data[k] == None:
                nums.append(0)

            elif (k == "scrape_date") and (type(data[k]) != float):
                stamp = datetime.strptime(data[k], "%Y-%m-%d %H:%M:%S.%f")
                epoch = datetime.utcfromtimestamp(0)
                nums.append((stamp - epoch).total_seconds())

            elif np.isreal(data[k]):
                nums.append(data[k])
        return nums