nearest-neighbor

问题 I'm new to machine learning and would like to setup a little sample using the k-nearest-Neighbor-method with the Python library Scikit . Transforming and fitting the data works fine but I can't figure out how to plot a graph showing the datapoints surrounded by their "neighborhood". The dataset I'm using looks like that: So there are 8 features, plus one "outcome" column. From my understanding, I get an array, showing the euclidean-distances of all datapoints, using the kneighbors_graph from

问题 I'm working with latitude and longitude data. I've used BallTree because I have many rows (32000 rows) in the dataset. If I built the tree with haversine distance: '''' model_BTree = BallTree(np.array(points_sec_rad),metric='haversine') '''' and I transform the latitude and longitude to radians units, how can I apply query_radius (max_dist_rad) to the points which I would like to locate? I've used 0.150 meters as radius but I'm not sure if I should use an approximation in radians. ''''ind

问题 I have two pandas dataframes I want to join/merge exactly on a number of columns (say 3) and approximately, i.e nearest neighbour, on one (date) column. I also want to return the difference (days) between them. Each dataset is about 50,000 rows long. I'm most interested in an inner join, but the “leftovers” are also interesting if not too hard to get hold of. Most of the “exact match” observations will exist multiple times in each data frame. I've been trying to use difflib.get_close_matches

问题 Python 3.4.3 (v3.4.3:9b73f1c3e601, Feb 23 2015, 02:52:03) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin Type "help", "copyright", "credits" or "license" for more information. import sklearn sklearn.__version__ '0.16.1' from sklearn.neighbors import NearestNeighbors Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages/sklearn/neighbors/__init__.py", line 9, in <module> from .graph

问题 Where can I find an serial C/C++ implementation of the k-nearest neighbour algorithm? Do you know of any library that has this? I have found openCV but the implementation is already parallel. I want to start from a serial implementation and parallelize it with pthreads openMP and MPI. Thanks, Alex 回答1: How about ANN? http://www.cs.umd.edu/~mount/ANN/. I have once used the kdtree implementation, but there are other options. Quoting from the website: "ANN is a library written in C++, which

问题 I'm first trying this out in R Studio with a small practice dataset found here (584 obs, 5 variables) (https://scholarworks.umass.edu/cgi/viewcontent.cgi?article=1330&context=pare) Using this code I can use nearest neighbor matching to find the mean difference in matched cases and controls (1:1) where stw is my grouping variable and tot, min, and dis are the matching variables: m.out = matchit(stw ~ tot + min + dis, data = mydata, method = "nearest", ratio = 1) what I want to know is how can

问题 I have 2 files: file1 -3241.42 633.261 1210.53 -1110.89 735.349 836.635 (this is the points I am looking for, with coordinates x,y,z) file2 2014124 -2277.576 742.75 962.5816 0 0 2036599 -3236.882 638.748 1207.804 0 0 2036600 -3242.417 635.2612 1212.527 0 0 2036601 -3248.006 631.6553 1217.297 0 0 2095885 -1141.905 737.7666 843.3465 0 0 2095886 -1111.889 738.3486 833.6354 0 0 2095887 -1172.227 737.4004 853.9965 0 0 2477149 -3060.679 488.6802 1367.816 0 0 2477150 -3068.369 489.6621 1365.769 0 0

问题 I have 2 files: file1 -3241.42 633.261 1210.53 -1110.89 735.349 836.635 (this is the points I am looking for, with coordinates x,y,z) file2 2014124 -2277.576 742.75 962.5816 0 0 2036599 -3236.882 638.748 1207.804 0 0 2036600 -3242.417 635.2612 1212.527 0 0 2036601 -3248.006 631.6553 1217.297 0 0 2095885 -1141.905 737.7666 843.3465 0 0 2095886 -1111.889 738.3486 833.6354 0 0 2095887 -1172.227 737.4004 853.9965 0 0 2477149 -3060.679 488.6802 1367.816 0 0 2477150 -3068.369 489.6621 1365.769 0 0

问题 Problem statement: I have 150k points in a 3D space with their coordinates stored in a matrix with dimension [150k, 3] in mm. I want to find all the neighbors of a given point p that are within a radius r . And I want to do that in the most accurate way. How should I choose my leafsize parameter ? from scipy.spatial import KDTree import numpy as np pts = np.random.rand(150000,3) T1 = KDTree(pts, leafsize=20) T2 = KDTree(pts, leafsize=1) neighbors1= T1.query_ball_point((0.3,0.2,0.1), r=2.0)

问题 Problem statement: I have 150k points in a 3D space with their coordinates stored in a matrix with dimension [150k, 3] in mm. I want to find all the neighbors of a given point p that are within a radius r . And I want to do that in the most accurate way. How should I choose my leafsize parameter ? from scipy.spatial import KDTree import numpy as np pts = np.random.rand(150000,3) T1 = KDTree(pts, leafsize=20) T2 = KDTree(pts, leafsize=1) neighbors1= T1.query_ball_point((0.3,0.2,0.1), r=2.0)