PCA vs averaging columns - machine-learning

I have a dataframe with 300 float type columns and 1 integer column which is the dependent variable. The 300 columns are of 3 kinds:
1.Kind A: columns 1 to 100
2.Kind B: columns 101 to 200
3.Kind C: columns 201 to 300
I want to reduce the number of dimensions. Should I average the values for each kind and aggregate into 3 columns(one for each type), or should I perform some dimensionality reduction techniques like PCA? What is the justification of the same?

Option 1:
Do not do dimensionality reduction if you have large number of training data (say more then 5*300 sample for training)
Option 2:
Since you know that there are 3 kinds of data, run a PCA of those three kinds separately and get say 2 features for each. i.e
f1, f2 = PCA(kind A columns)
f3, f4 = PCA(kind B columns)
f5, f6 = PCA(kind C columns)
train(f1, f2, f3, f4, f5, f6)
Option 3
Run PCA on all columns and only take number of columns which preserve 90+ variance
Do not average, averaging is bad. But if you really want to do averaging and if you know certainly that some features are important rather do weighted average. In general averaging of features for dimensionally reduction is a very bad idea.

PCA will only consider the rows which will have highest co-relation with the output / result. So not all rows will be considered as a part of process to determine the output.
So it will be better if u do averaging as it will consider all the rows and determine the output from them.
As u have a larger number of features it is better if all the features are used to determine output.

Related

Is F1 score a good measure for balanced dataset

My dataset is approxiately balanced: 52/48. I evaluate both ACC and F1-score. The result returned by Random forest model is below
Acc: 52%
F1: 68%
Confusion matrix:
|Predicted
Label|0 |1
0 |52|122109
1 |19|134802
I know if I switch labels 0 as 1 and vice versa, the F1 score will be very small.
So, in the case of using F1, should I always switch labels?
The interpretation of F1 score entirely depends on an arbitrary choice of labels (this is buried within its formulation). F1 score is therefore most suitable for cases where class labels actually mean and correspond to negative and positive in real-life (e.g., presence of cancer) and where there is an imbalanced class distribution (particularly, when the negatives significantly outnumber the positives). Since your data is balanced and it seems you can arbitrarily switch the labels too, F1 score may not be a suitable metric to use.
These may also help: 1 2 3

How to squish a continuous cosine-theta score to a discrete (0/1) output?

I implemented a cosine-theta function, which calculates the relation between two articles. If two articles are very similar then the words should contain quite some overlap. However, a cosine theta score of 0.54 does not mean "related" or "not related". I should end up with a definitive answer which is either 0 for 'not related' or 1 for 'related'.
I know that there are sigmoid and softmax functions, yet I should find the optimal parameters to give to such functions and I do not know if these functions are satisfactory solutions. I was thinking that I have the cosine theta score, I can calculate the percentage of overlap between two sentences two (e.g. the amount of overlapping words divided by the amount of words in the article) and maybe some more interesting things. Then with the data, I could maybe write a function (what type of function I do not know and is part of the question!), after which I can minimize the error via the SciPy library. This means that I should do some sort of supervised learning, and I am willing to label article pairs with labels (0/1) in order to train a network. Is this worth the effort?
# Count words of two strings.
v1, v2 = self.word_count(s1), self.word_count(s2)
# Calculate the intersection of the words in both strings.
v3 = set(v1.keys()) & set(v2.keys())
# Calculate some sort of ratio between the overlap and the
# article length (since 1 overlapping word on 2 words is more important
# then 4 overlapping words on articles of 492 words).
p = min(len(v1), len(v2)) / len(v3)
numerator = sum([v1[w] * v2[w] for w in v3])
w1 = sum([v1[w]**2 for w in v1.keys()])
w2 = sum([v2[w]**2 for w in v2.keys()])
denominator = math.sqrt(w1) * math.sqrt(w2)
# Calculate the cosine similarity
if not denominator:
return 0.0
else:
return (float(numerator) / denominator)
As said, I would like to use variables such as p, and the cosine theta score in order to produce an accurate discrete binary label, either 0 or 1.
As said, I would like to use variables such as p, and the cosine theta score in order to produce an accurate discrete binary label, either 0 or 1.
Here it really comes down to what you mean by accuracy. It is up to you to choose how the overlap affects whether or not two strings are "matching" unless you have a labelled data set. If you have a labelled data set (I.e., a set of pairs of strings along with a 0 or 1 label), then you can train a binary classification algorithm and try to optimise based on that. I would recommend something like a neural net or SVM due to the potentially high dimensional, categorical nature of your problem.
Even the optimisation, however, is a subjective measure. For example, in theory let's pretend you have a model which out of 100 samples only predicts 1 answer (Giving 99 unknowns). Technically if that one answer is correct, that is a model with 100% accuracy, but which has a very low recall. Generally in machine learning you will find a trade off between recall and accuracy.
Some people like to go for certain metrics which combine the two (The most famous of which is the F1 score), but honestly it depends on the application. If I have a marketing campaign with a fixed budget, then I care more about accuracy - I would only want to target consumers who are likely to buy my product. If however, we are looking to test for a deadly disease or markers for bank fraud, then it's feasible for that test to be accurate only 10% of the time - if its recall of true positives is somewhere close to 100%.
Finally, if you have no labelled data, then your best bet is just to define some cut off value which you believe indicates a good match. This is would then be more analogous to a binary clustering problem, and you could use some more abstract measure such as distance to a centroid to test which cluster (Either the "related" or "unrelated" cluster) the point belongs to. Note however that here your features feel like they would be incredibly hard to define.

How to calculate the accuracy of classes from a 7x7 confusion matrix?

So I've got the following results from Naïves Bayes classification on my data set:
I am stuck however on understanding how to interpret the data. I am wanting to find and compare the accuracy of each class (a-g).
I know accuracy is found using this formula:
However, lets take the class a. If I take the number of correctly classified instances - 313 - and divide it by the total number of 'a' (4953) from the row a, this gives ~6.32%. Would this be the accuracy?
EDIT: if we use the column instead of the row, we get 313/1199 which gives ~26.1% which seems a more reasonable number.
EDIT 2: I have done a calculation of the accuracy of a in excel which gives me 84% as the accuracy, using the accuracy calculation shown above:
This doesn't seem right, as the overall accuracy of classification successfully is ~24%
No -- all you've calculated is tp/(tp+fn), the total correct identifications of class a, divided by the total of actual a examples. This is recall, not accuracy. You need to include the other two figures.
fp is the rest of the a column; tn is all of the other figures in the non-a rows and columns, the 6x6 sub-matrix. This will reduce all 35K+ trials to a 2x2 matrix with labels a and not a, the 2x2 confusion matrix with which you're already familiar.
Yes, you get to repeat that reduction for each of the seven features. I recommend doing it programmatically.
RESPONSE TO OP UPDATE
Your accuracy is that high: you have a huge quantity of true negatives, not-a samples that were properly classified as not-a.
Perhaps it doesn't feel right because our experience focuses more on the class in question. There are [other statistics that handle that focus.
Recall is tp / (tp+fn) -- of all items actually in class a, what percentage did we properly identify? This is the 6.32% figure.
Precision is tp / (tp + fp) -- of all items identified as class a, what percentage were actually in that class. This is the 26.1% figure you calculated.

k-means for all data or for each feature?

I want use k-means to discretize a time series data in two values (0 or 1). My time series data is a matrix time per genes (line = time, column = gene). Ex:
t\x x1 x2 x3
1 0.122 0.324 0.723
2 0.543 0.573 0.329
3 0.901 0.445 0.343
4 0.612 0.353 0.435
5 0.192 0.233 0.023
My question: Should I use k clusters for all data of matrix or k clusters for each column (so I will have k cluster per column totalizing k.number_columns)? and my genes are independents
Either may work.
Discretising all attributes at once has the benefit of giving you only one symbol per time, i.e. a univariate series.
But on the other hand, if columns are independent, the quality could be better if you discretise them individually. Note thatfor one-dimensional data, if it is noisy, quantiles may be much better than k-means (which is sensitive to noise).

Genetic algorithms: fitness function for feature selection algorithm

I have data set n x m where there are n observations and each observation consists of m values for m attributes. Each observation has also observed result assigned to it. m is big, too big for my task. I am trying to find a best and smallest subset of m attributes that still represents the whole dataset quite well, so that I could use only these attributes for teaching a neural network.
I want to use genetic algorithm for this. The problem is the fittness function. It should tell how well the generated model (subset of attributes) still reflects the original data. And I don't know how to evaluate certain subset of attributes against the whole set.
Of course I could use the neural network(that will later use this selected data anyway) for checking how good the subset is - the smaller the error, the better the subset. BUT, this takes a looot of time in my case and I do not want to use this solution. I am looking for some other way that would preferably operate only on the data set.
What I thought about was: having subset S (found by genetic algorithm), trim data set so that it contains values only for subset S and check how many observations in this data ser are no longer distinguishable (have same values for same attributes) while having different result values. The bigger the number is, the worse subset it is. But this seems to me like a bit too computationally exhausting.
Are there any other ways to evaluate how well a subset of attributes still represents the whole data set?
This cost function should do what you want: sum the factor loadings that correspond to the features comprising each subset.
The higher that sum, the greater the share of variability in the response variable that is explained with just those features. If i understand the OP, this cost function is a faithful translation of "represents the whole set quite well" from the OP.
Reducing to code is straightforward:
Calculate the covariance matrix of your dataset (first remove the
column that holds the response variable, i.e., probably the last
one). If your dataset is m x n (columns x rows), then this
covariance matrix will be n x n, with "1"s down the main diagonal.
Next, perform an eigenvalue decomposition on this covariance
matrix; this will give you the proportion of the total variability
in the response variable, contributed by that eigenvalue (each
eigenvalue corresponds to a feature, or column). [Note,
singular-value decomposition (SVD) is often used for this step, but
it's unnecessary--an eigenvalue decomposition is much simpler, and
always does the job as long as your matrix is square, which
covariance matrices always are].
Your genetic algorithm will, at each iteration, return a set of
candidate solutions (features subsets, in your case). The next task
in GA, or any combinatorial optimization, is to rank those candiate
solutions by their cost function score. In your case, the cost
function is a simple summation of the eigenvalue proportion for each
feature in that subset. (I guess you would want to scale/normalize
that calculation so that the higher numbers are the least fit
though.)
A sample calculation (using python + NumPy):
>>> # there are many ways to do an eigenvalue decomp, this is just one way
>>> import numpy as NP
>>> import numpy.linalg as LA
>>> # calculate covariance matrix of the data set (leaving out response variable column)
>>> C = NP.corrcoef(d3, rowvar=0)
>>> C.shape
(4, 4)
>>> C
array([[ 1. , -0.11, 0.87, 0.82],
[-0.11, 1. , -0.42, -0.36],
[ 0.87, -0.42, 1. , 0.96],
[ 0.82, -0.36, 0.96, 1. ]])
>>> # now calculate eigenvalues & eivenvectors of the covariance matrix:
>>> eva, evc = LA.eig(C)
>>> # now just get value proprtions of each eigenvalue:
>>> # first, sort the eigenvalues, highest to lowest:
>>> eva1 = NP.sort(eva)[::-1]
>>> # get value proportion of each eigenvalue:
>>> eva2 = NP.cumsum(eva1/NP.sum(eva1)) # "cumsum" is just cumulative sum
>>> title1 = "ev value proportion"
>>> print( "{0}".format("-"*len(title1)) )
-------------------
>>> for row in q :
print("{0:1d} {1:3f} {2:3f}".format(int(row[0]), row[1], row[2]))
ev value proportion
1 2.91 0.727
2 0.92 0.953
3 0.14 0.995
4 0.02 1.000
so it's the third column of values just above (one for each feature) that are summed (selectively, depending on which features are present in a given subset you are evaluating with the cost function).

Resources