geo_median.py 文件源码

def compute_user_median(data_points, num_iter, csvwriter, current_uid):
    if len(data_points) < LIMIT_POINTS:  # Insufficient points for the user - don't record median
        if OUTPUT_ALL_USERS:
            csvwriter.writerow([current_uid, None])
    else:
        if SNAP_TO_USER_POINTS: # ensure median is one of the user's points
            lowest_dev = float("inf")
            for point in data_points:
                tmp_abs_dev = objfunc(point, data_points)
                if tmp_abs_dev < lowest_dev:
                    lowest_dev = tmp_abs_dev
                    test_median = point
        else:
            test_median = cand_median(data_points)  # Calculate centroid more or less as starting point
            if objfunc(test_median, data_points) != 0:  # points aren't all the same
                # iterate to find reasonable estimate of median
                for x in range(0, num_iter):
                    denom = denomsum(test_median, data_points)
                    next_lat = 0.0
                    next_lon = 0.0

                    for y in range(0, len(data_points)):
                        next_lat += (data_points[y][0] * numersum(test_median, data_points[y])) / denom
                        next_lon += (data_points[y][1] * numersum(test_median, data_points[y])) / denom

                    prev_median = test_median
                    test_median = (next_lat, next_lon)
                    try:
                        if vincenty(prev_median, test_median).meters < DISTANCE_THRESHOLD:
                            break
                    except:
                        if great_circle(prev_median, test_median).meters < DISTANCE_THRESHOLD:
                            break

                if x == num_iter - 1:
                    print('{0}: failed to converge. Last change between iterations was {1} meters.'.format(current_uid, great_circle(prev_median, test_median).meters))

        # Check if user points are under the limit median absolute deviation
        if check_median_absolute_deviation(data_points, test_median) <= LIMIT_MAD:
            csvwriter.writerow([current_uid, (round(test_median[0],6), round(test_median[1],6))])
        else:
            if OUTPUT_ALL_USERS:
                csvwriter.writerow([current_uid, None])