Data Manipulation¶
H2OFrame¶
-
class
h2o.frame.H2OFrame(python_obj=None)[source]¶ Bases:
object-
anyfactor()[source]¶ Test if H2OFrame has any factor columns.
Returns: True if there are any categorical columns; False otherwise.
-
apply(fun=None, axis=0)[source]¶ Apply a lambda expression to an H2OFrame.
Parameters: fun: lambda
A lambda expression to be applied per row or per column
- axis: int
0: apply to each column; 1: apply to each row
Returns: H2OFrame
-
as_data_frame(use_pandas=True)[source]¶ Obtain the dataset as a python-local object.
Parameters: use_pandas : bool, default=True
A flag specifying whether or not to return a pandas DataFrame.
Returns: A local python object (a list of lists of strings, each list is a row, if
use_pandas=False, otherwise a pandas DataFrame) containing this H2OFrame instance’s data.
-
as_date(format)[source]¶ Return the column with all elements converted to millis since the epoch.
Parameters: format : str
A datetime format string (e.g. “YYYY-mm-dd”)
Returns: An H2OFrame instance.
-
cbind(data)[source]¶ Append data to this H2OFrame column-wise.
Parameters: data : H2OFrame
H2OFrame to be column bound to the right of this H2OFrame.
Returns: H2OFrame of the combined datasets.
-
col_names¶ Returns: A list of column names.
-
columns¶ Returns: A list of column names.
-
concat(frames, axis=1)[source]¶ Append multiple data to this H2OFrame column-wise
Parameters: frames : List of H2OFrame’s
H2OFrame’s to be column bound to the right of this H2OFrame.
- axis : int, default = 1
Type of concatenation to conduct. If axis = 1, then column-wise (Default). If axis = 0, then row-wise.
Returns: H2OFrame of the combined datasets.
-
cor(y=None, na_rm=False, use=None)[source]¶ Compute the correlation matrix of one or two H2OFrames.
Parameters: y : H2OFrame, default=None
If y is None and self is a single column, then the correlation is computed for self. If self has multiple columns, then its correlation matrix is returned. Single rows are treated as single columns. If y is not None, then a correlation matrix between the columns of self and the columns of y is computed.
na_rm : bool, default=False
Remove NAs from the computation.
use : str, default=None, which acts as “everything” if na_rm is False, and “complete.obs” if na_rm is True
- A string indicating how to handle missing values. This must be one of the following:
“everything” - outputs NaNs whenever one of its contributing observations is missing “all.obs” - presence of missing observations will throw an error “complete.obs” - discards missing values along with all observations in their rows so that only complete observations are used
Returns
——-
An H2OFrame of the correlation matrix of the columns of this H2OFrame with itself (if y is not given), or with the columns of y (if y is given). If self and y are single rows or single columns, the correlation is given as a scalar.
-
countmatches(pattern)[source]¶ For each string in the column, count the occurrences of pattern.
Parameters: pattern : str
The pattern to count matches on in each string.
Returns: A single-column H2OFrame containing the counts for the per-row occurrences of
pattern in the input column.
-
cut(breaks, labels=None, include_lowest=False, right=True, dig_lab=3)[source]¶ Cut a numeric vector into factor “buckets”. Similar to R’s cut method.
Parameters: breaks : list
The cut points in the numeric vector (must span the range of the col.)
- labels: list
Factor labels, defaults to set notation of intervals defined by breaks.
- include_lowest : bool
By default, cuts are defined as (lo,hi]. If True, get [lo,hi].
- right : bool
Include the high value: (lo,hi]. If False, get (lo,hi).
- dig_lab: int
Number of digits following the decimal point to consider.
Returns: Single-column H2OFrame of categorical data.
-
describe()[source]¶ Generate an in-depth description of this H2OFrame. Everything in summary(), plus the data layout.
-
diff()[source]¶ Computes the lag1 diff on a numeric column.
Returns: The lag1 difference for a numeric column (expects operation to occur over H2OFrame
of a single column).
-
dim¶ Returns: The number of rows and columns in the H2OFrame as a list [rows, cols].
-
drop(i)[source]¶ Drop a column from the current H2OFrame.
Parameters: i : str, int
The column to be dropped
Returns: H2OFrame with the column at index i dropped. Returns a new H2OFrame.
-
entropy()[source]¶ For each string, return the Shannon entropy. If the string is empty, the entropy is 0.
Returns: An H2OFrame of Shannon entropies.
-
filter_na_cols(frac=0.2)[source]¶ Filter columns with proportion of NAs >= frac.
Parameters: frac : float
Fraction of NAs in the column.
Returns: A list of column indices that have a fewer count of NAs.
If all columns are filtered, None is returned.
-
frame_id¶ Returns: Get the name of this frame.
-
static
from_python(python_obj, destination_frame='', header=(-1, 0, 1), separator='', column_names=None, column_types=None, na_strings=None)[source]¶ Properly handle native python data types. For a discussion of the rules and permissible data types please refer to the main documentation for H2OFrame.
Parameters: python_obj : tuple, list, dict, collections.OrderedDict
If a nested list/tuple, then each nested collection is a row.
- destination_frame : str, optional
The unique hex key assigned to the imported file. If none is given, a key will automatically be generated.
- header : int, optional
-1 means the first line is data, 0 means guess, 1 means first line is header.
- sep : str, optional
The field separator character. Values on each line of the file are separated by this character. If sep = “”, the parser will automatically detect the separator.
- col_names : list, optional
A list of column names for the file.
- col_types : list or dict, optional
A list of types or a dictionary of column names to types to specify whether columns should be forced to a certain type upon import parsing. If a list, the types for elements that are None will be guessed. The possible types a column may have are.
- na_strings : list or dict, optional
A list of strings, or a list of lists of strings (one list per column), or a dictionary of column names to strings which are to be interpreted as missing values.
Returns: A new H2OFrame instance.
Examples
>>> l = [[1,2,3,4,5], [99,123,51233,321]] >>> l = H2OFrame(l) >>> l
-
static
get_frame(frame_id)[source]¶ Create an H2OFrame mapped to an existing id in the cluster.
Returns: H2OFrame that points to a pre-existing big data H2OFrame in the cluster
-
get_frame_data()[source]¶ Get frame data as str in csv format
Returns: A local python string, each line is a row and each element separated by commas,
containing this H2OFrame instance’s data.
-
group_by(by)[source]¶ - Returns a new GroupBy object using this frame and the desired grouping columns.
- The returned groups are sorted by the natural group-by column sort.
Parameters: by : list
The columns to group on.
Returns: A new GroupBy object.
-
gsub(pattern, replacement, ignore_case=False)[source]¶ Globally substitute occurrences of pattern in a string with replacement.
Parameters: pattern : str
A regular expression.
- replacement : str
A replacement string.
- ignore_case : bool
If True then pattern will match against upper and lower case.
Returns: H2OFrame
-
head(rows=10, cols=200)[source]¶ Analogous to Rs head call on a data.frame.
Parameters: rows : int, default=10
Number of rows starting from the topmost
- cols : int, default=200
Number of columns starting from the leftmost
Returns: An H2OFrame.
-
hist(breaks='Sturges', plot=True, **kwargs)[source]¶ Compute a histogram over a numeric column.
Parameters: breaks: str, int, list
Can be one of “Sturges”, “Rice”, “sqrt”, “Doane”, “FD”, “Scott.” Can be a single number for the number of breaks. Can be a list containing sthe split points, e.g., [-50,213.2123,9324834] If breaks is “FD”, the MAD is used over the IQR in computing bin width.
- plot : bool, default=True
If True, then a plot is generated
Returns: If plot is False, return H2OFrame with these columns: breaks, counts, mids_true,
mids, and density; otherwise produce the plot.
-
ifelse(yes, no)[source]¶ Equivalent to [y if t else n for t,y,n in zip(self,yes,no)]
Based on the booleans in the test vector, the output has the values of the yes and no vectors interleaved (or merged together). All Frames must have the same row count. Single column frames are broadened to match wider Frames. Scalars are allowed, and are also broadened to match wider frames.
Parameters: test : H2OFrame (self)
Frame of values treated as booleans; may be a single column
- yes : H2OFrame
Frame to use if [test] is true ; may be a scalar or single column
- no : H2OFrame
Frame to use if [test] is false; may be a scalar or single column
Returns: H2OFrame of the merged yes/no Frames/scalars according to the test input frame.
-
impute(column=-1, method='mean', combine_method='interpolate', by=None, group_by_frame=None, values=None)[source]¶ Impute in place.
Parameters: column: int, default=-1
The column to impute, if -1 then impute the whole frame
- method : str, default=”mean”
The method of imputation: mean, median, mode
- combine_method : str, default=”interpolate”
When method is “median”, dictates how to combine quantiles for even samples.
- by : list, default=None
The columns to group on.
- group_by_frame : H2OFrame, default=None
Impute the column col with this pre-computed grouped frame.
- values : list
A list of impute values (one per column). NaN indicates to skip the column.
Returns: A list of values used in the imputation or the group by result used in imputation.
-
insert_missing_values(fraction=0.1, seed=None)[source]¶ Inserting Missing Values into an H2OFrame. Randomly replaces a user-specified fraction of entries in a H2O dataset with missing values.
WARNING! This will modify the original dataset. Unless this is intended, this function should only be called on a subset of the original.
Parameters: fraction : float
A number between 0 and 1 indicating the fraction of entries to replace with missing.
- seed : int
A random number used to select which entries to replace with missing values.
Returns: H2OFrame with missing values inserted.
-
interaction(factors, pairwise, max_factors, min_occurrence, destination_frame=None)[source]¶ Categorical Interaction Feature Creation in H2O. Creates a frame in H2O with n-th order interaction features between categorical columns, as specified by the user.
Parameters: factors : list
factors Factor columns (either indices or column names).
- pairwise : bool
Whether to create pairwise interactions between factors (otherwise create one higher-order interaction). Only applicable if there are 3 or more factors.
- max_factors: int
Max. number of factor levels in pair-wise interaction terms (if enforced, one extra catch-all factor will be made)
- min_occurrence: int
Min. occurrence threshold for factor levels in pair-wise interaction terms
- destination_frame: str, optional
A string indicating the destination key.
Returns: H2OFrame
-
ischaracter()[source]¶ Returns: True if the column is a character column, otherwise False (same as isstring)
-
isfactor()[source]¶ Test if the selection is a factor column.
Returns: True if the column is categorical; otherwise False. For String columns, the result
is False.
-
isin(item)[source]¶ Test whether elements of an H2OFrame are contained in the item.
Parameters: items : any element or a list of elements
An item or a list of items to compare the H2OFrame against.
Returns: An H2OFrame of 0s and 1s showing whether each element in the original H2OFrame is contained in item.
-
isna()[source]¶ For each element in an H2OFrame, determine if it is NA or not.
Returns: H2OFrame of 1s and 0s. 1 means the value was NA.
-
isstring()[source]¶ Returns: True if the column is a string column, otherwise False (same as ischaracter)
-
kfold_column(n_folds=3, seed=-1)[source]¶ Build a fold assignments column for cross-validation. This call will produce a column having the same data layout as the calling object.
Parameters: n_folds : int
An integer specifying the number of validation sets to split the training data into.
- seed : int, optional
Seed for random numbers as fold IDs are randomly assigned.
Returns: A single column H2OFrame with the fold assignments.
-
lstrip(set=' ')[source]¶ Return a copy of the column with leading characters removed. The set argument is a string specifying the set of characters to be removed. If omitted, the set argument defaults to removing whitespace.
Parameters: set : str
Set of characters to lstrip from strings in column
Returns: H2OFrame with lstripped strings.
-
match(table, nomatch=0)[source]¶ Makes a vector of the positions of (first) matches of its first argument in its second.
Parameters: table : list
list of items to match against
nomatch : optional
Returns: H2OFrame of one boolean column
-
mean(na_rm=False)[source]¶ Compute the mean.
Parameters: na_rm: bool, default=False
If True, then remove NAs from the computation.
Returns: A list containing the mean for each column (NaN for non-numeric columns).
-
median(na_rm=False)[source]¶ Compute the median.
Parameters: na_rm: bool, default=False
If True, then remove NAs from the computation.
Returns: A list containing the median for each column (NaN for non-numeric columns).
-
merge(other, all_x=False, all_y=False, by_x=None, by_y=None, method='auto')[source]¶ Merge two datasets based on common column names
Parameters: other: H2OFrame
Other dataset to merge. Must have at least one column in common with self, and all columns in common are used as the merge key. If you want to use only a subset of the columns in common, rename the other columns so the columns are unique in the merged result.
- all_x: bool, default=False
If True, include all rows from the left/self frame
- all_y: bool, default=False
If True, include all rows from the right/other frame
Returns: Original self frame enhanced with merged columns and rows
-
static
mktime(year=1970, month=0, day=0, hour=0, minute=0, second=0, msec=0)[source]¶ All units are zero-based (including months and days). Missing year is 1970.
Parameters: year : int, H2OFrame
the year
- month: int, H2OFrame
the month
- day : int, H2OFrame
the day
- hour : int, H2OFrame
the hour
- minute : int, H2OFrame
the minute
- second : int, H2OFrame
the second
- msec : int, H2OFrame
the milisecond
Returns: H2OFrame of one column containing the date in millis since the epoch.
-
modulo_kfold_column(n_folds=3)[source]¶ Build a fold assignments column for cross-validation. Rows are assigned a fold according to the current row number modulo n_folds.
Parameters: n_folds : int
An integer specifying the number of validation sets to split the training data into.
Returns: A single column H2OFrame with the fold assignments.
-
mult(matrix)[source]¶ Perform matrix multiplication.
Parameters: matrix : H2OFrame
The right-hand-side matrix
Returns: H2OFrame result of the matrix multiplication
-
nacnt()[source]¶ Count of NAs for each column in this H2OFrame.
Returns: A list of the na cnts (one entry per column).
-
names¶ Retrieve the column names (one name per H2OVec) for this H2OFrame.
Returns: A str list of column names
-
nchar()[source]¶ Count the number of characters in each string of single-column H2OFrame.
Returns: A single-column H2OFrame containing the per-row character count.
-
ncol¶ Returns: The number of columns in the H2OFrame.
-
nlevels()[source]¶ Get the number of factor levels for this frame.
Returns: A list of the number of levels per column.
-
nrow¶ Returns: The number of rows in the H2OFrame.
-
num_valid_substrings(path_to_words)[source]¶ For each string, find the count of all possible substrings >= 2 characters that are contained in the line-separated text file whose path is given.
Parameters: path_to_words : str
Path to file that contains a line-separated list of strings considered valid.
Returns: An H2OFrame with the number of substrings that are contained in the given word list.
-
pop(i)[source]¶ Pop a column from the H2OFrame at index i
Parameters: i : int, str
The index or name of the column to pop.
Returns: The column dropped from the frame; the frame is side-effected to lose the column.
-
prod(na_rm=False)[source]¶ Parameters: na_rm : bool, default=False
True or False to remove NAs from computation.
Returns: The product of the column.
-
quantile(prob=None, combine_method='interpolate', weights_column=None)[source]¶ Compute quantiles.
Parameters: prob : list, default=[0.01,0.1,0.25,0.333,0.5,0.667,0.75,0.9,0.99]
A list of probabilities of any length.
- combine_method : str, default=”interpolate”
For even samples, how to combine quantiles. Should be one of [“interpolate”, “average”, “low”, “high”]
- weights_column : str, default=None
Name of column with optional observation weights in this H2OFrame or a 1-column H2OFrame of observation weights.
Returns: A new H2OFrame containing the quantiles and probabilities.
-
rbind(data)[source]¶ Combine H2O Datasets by rows. Takes a sequence of H2O data sets and combines them by rows.
Parameters: data : H2OFrame Returns: Returns this H2OFrame with data appended row-wise.
-
relevel(y)[source]¶ Reorders levels of an H2O factor, similarly to standard R’s relevel(). The levels of a factor are reordered such that the reference level is at level 0, remaining levels are moved down as needed.
Parameters: x: Column
Column in H2O Frame
- y : String
Reference level
Returns: New reordered factor column
-
rep_len(length_out)[source]¶ Replicates the values in data in the H2O backend
Parameters: length_out : int
Number of columns of the resulting H2OFrame
Returns: H2OFrame
-
round(digits=0)[source]¶ Round doubles/floats to the given number of decimal places.
Parameters: digits : int, default=0
Number of decimal places to round doubles/floats. Rounding to a negative number of decimal places is not supported. For rounding off a 5, the IEC 60559 standard is used, ‘go to the even digit’. Therefore rounding 2.5 gives 2 and rounding 3.5 gives 4.
Returns: H2OFrame
-
rstrip(set=' ')[source]¶ Return a copy of the column with trailing characters removed. The set argument is a string specifying the set of characters to be removed. If omitted, the set argument defaults to removing whitespace.
Parameters: set : str
Set of characters to rstrip from strings in column
Returns: H2OFrame with rstripped strings.
-
runif(seed=None)[source]¶ Generate a column of random numbers drawn from a uniform distribution [0,1) and having the same data layout as the calling H2OFrame instance.
Parameters: seed : int, optional
A random seed. If None, then one will be generated.
Returns: Single-column H2OFrame filled with doubles sampled uniformly from [0,1).
-
scale(center=True, scale=True)[source]¶ Centers and/or scales the columns of the self._newExpr
Parameters: center : bool, list
If True, then demean the data by the mean. If False, no shifting is done. If a list, then shift each column by the given amount in the list.
- scale : bool, list
If True, then scale the data by the column standard deviation. If False, no scaling is done. If a list, then scale each column by the given amount in the list.
Returns: H2OFrame
-
sd(na_rm=False)[source]¶ Compute the standard deviation.
Parameters: na_rm : bool, default=False
Remove NAs from the computation.
Returns: A list containing the standard deviation for each column (NaN for non-numeric
columns).
-
set_level(level)[source]¶ A method to set all column values to one of the levels.
Parameters: level : str
The level at which the column will be set (a string)
Returns: H2OFrame with entries set to the desired level.
-
set_levels(levels)[source]¶ Works on a single categorical column. New domains must be aligned with the old domains. This call has copy-on-write semantics.
Parameters: levels : list
A list of strings specifying the new levels. The number of new levels must match the number of old levels.
Returns: A single-column H2OFrame with the desired levels.
-
set_name(col=None, name=None)[source]¶ Set the name of the column at the specified index.
Parameters: col : int, str
Index of the column whose name is to be set; may be skipped for 1-column frames
- name : str
The new name of the column to set
Returns: Returns self.
-
set_names(names)[source]¶ Change all of this H2OFrame instance’s column names.
Parameters: names : list
A list of strings equal to the number of columns in the H2OFrame.
-
shape¶ Returns: A tuple (nrow, ncol)
-
show(use_pandas=False)[source]¶ Used by the H2OFrame.__repr__ method to print or display a snippet of the data frame. If called from IPython, displays an html’ized result Else prints a tabulate’d result
-
signif(digits=6)[source]¶ Round doubles/floats to the given number of significant digits.
Parameters: digits : int, default=6
Number of significant digits to round doubles/floats.
Returns: H2OFrame
-
split_frame(ratios=None, destination_frames=None, seed=None)[source]¶ Split a frame into distinct subsets of size determined by the given ratios. The number of subsets is always 1 more than the number of ratios given. Note that this does not give an exact split. H2O is designed to be efficient on big data using a probabilistic splitting method rather than an exact split. For example when specifying a split of 0.75/0.25, H2O will produce a test/train split with an expected value of 0.75/0.25 rather than exactly 0.75/0.25. On small datasets, the sizes of the resulting splits will deviate from the expected value more than on big data, where they will be very close to exact.
Parameters: ratios : list
The fraction of rows for each split.
- destination_frames : list
The names of the split frames.
- seed : int
Used for selecting which H2OFrame a row will belong to.
Returns: A list of H2OFrame instances
-
stratified_kfold_column(n_folds=3, seed=-1)[source]¶ Build a fold assignment column with the constraint that each fold has the same class distribution as the fold column.
Parameters: n_folds: int
The number of folds to build.
- seed: int
A random seed.
Returns: A single column H2OFrame with the fold assignments.
-
stratified_split(test_frac=0.2, seed=-1)[source]¶ Construct a column that can be used to perform a random stratified split.
Parameters: test_frac : float, default=0.2
The fraction of rows that will belong to the “test”.
- seed : int
For seeding the random splitting.
Returns: A categorical column of two levels “train” and “test”.
Examples
>>> my_stratified_split = my_frame["response"].stratified_split(test_frac=0.3,seed=12349453) >>> train = my_frame[my_stratified_split=="train"] >>> test = my_frame[my_stratified_split=="test"]
# check the distributions among the initial frame, and the train/test frames match >>> my_frame[“response”].table()[“Count”] / my_frame[“response”].table()[“Count”].sum() >>> train[“response”].table()[“Count”] / train[“response”].table()[“Count”].sum() >>> test[“response”].table()[“Count”] / test[“response”].table()[“Count”].sum()
-
strsplit(pattern)[source]¶ Split the strings in the target column on the given pattern
Parameters: pattern : str
The split pattern.
Returns: H2OFrame containing columns of the split strings.
-
sub(pattern, replacement, ignore_case=False)[source]¶ Substitute the first occurrence of pattern in a string with replacement.
Parameters: pattern : str
A regular expression.
- replacement : str
A replacement string.
- ignore_case : bool
If True then pattern will match against upper and lower case.
Returns: H2OFrame
-
substring(start_index, end_index=None)[source]¶ For each string, return a new string that is a substring of the original string. If end_index is not specified, then the substring extends to the end of the original string. If the start_index is longer than the length of the string, or is greater than or equal to the end_index, an empty string is returned. Negative start_index is coerced to 0.
Parameters: start_index : int
The index of the original string at which to start the substring, inclusive.
- end_index: int, optional
The index of the original string at which to end the substring, exclusive.
Returns: An H2OFrame containing the specified substrings.
-
table(data2=None, dense=True)[source]¶ Compute the counts of values appearing in a column, or co-occurence counts between two columns.
Parameters: data2 : H2OFrame
Default is None, can be an optional single column to aggregate counts by.
- dense : bool
Default is True, for dense representation, which lists only non-zero counts, 1 combination per row. Set to False to expand counts across all combinations.
Returns: H2OFrame of the counts at each combination of factor levels
-
tail(rows=10, cols=200)[source]¶ Analogous to Rs tail call on a data.frame.
Parameters: rows : int, default=10
Number of rows starting from the bottommost
- cols: int, default=200
Number of columns starting from the leftmost
Returns: An H2OFrame.
-
tolower()[source]¶ Translate characters from upper to lower case for a particular column
Returns: H2OFrame
-
toupper()[source]¶ Translate characters from lower to upper case for a particular column
Returns: H2OFrame
-
transpose()[source]¶ Transpose rows and columns of H2OFrame.
Returns: The transpose of the input frame.
-
trim()[source]¶ Trim white space on the left and right of strings in a single-column H2OFrame.
Returns: H2OFrame with trimmed strings.
-
types¶ Returns: A dictionary of column_name-type pairs.
-
unique()[source]¶ Extract the unique values in the column.
Returns: H2OFrame of just the unique values in the column.
-
var(y=None, na_rm=False, use=None)[source]¶ Compute the variance or covariance matrix of one or two H2OFrames. Parameters ———- y : H2OFrame, default=None
If y is None and self is a single column, then the variance is computed for self. If self has multiple columns, then its covariance matrix is returned. Single rows are treated as single columns. If y is not None, then a covariance matrix between the columns of self and the columns of y is computed.- na_rm : bool, default=False
- Remove NAs from the computation.
- use : str, default=None, which acts as “everything” if na_rm is False, and “complete.obs” if na_rm is True
- A string indicating how to handle missing values. This must be one of the following:
- “everything” - outputs NaNs whenever one of its contributing observations is missing “all.obs” - presence of missing observations will throw an error “complete.obs” - discards missing values along with all observations in their rows so that only complete observations are used
An H2OFrame of the covariance matrix of the columns of this H2OFrame with itself (if y is not given), or with the columns of y (if y is given). If self and y are single rows or single columns, the variance or covariance is given as a scalar.
-
GroupBy¶
-
class
h2o.group_by.GroupBy(fr, by)[source]¶ A class that represents the group by operation on an H2OFrame.
Sample usage:
>>> my_frame = ... # some existing H2OFrame >>> grouped = my_frame.group_by(by=["C1","C2"]) >>> grouped.sum(col="X1",na="all").mean(col="X5",na="all").max() >>> grouped.get_frame()
Any number of aggregations may be chained together in this manner.
If no arguments are given to the aggregation (e.g. “max” in the above example), then it is assumed that the aggregation should apply to all columns but the group by columns.
- The na parameter is one of [“all”,”ignore”,”rm”].
- “all” - include NAs “rm” - exclude NAs
Variance (var) and standard deviation (sd) are the sample (not population) statistics.
-
frame¶ Returns: the result of the group by