Questions tagged [feature-engineering]
the process of using domain knowledge of the data to create features that improve machine learning algorithms
309
questions
3
votes
1answer
31 views
Overfitting due to features correlating with training set generation rules
As background, I am using a Deep Neural Network built using Keras to classify inputs into 5 categories.
The current structure of the network is:
Input layer (~450 nodes)
Dense layer (750 nodes)
...
1
vote
1answer
36 views
How can we convert time series data to supervised learning problem?
I am preparing a data for machine learning model. I want to deal with time series data as normal supervised learning prediction. Let's say I have a data for car speed and I have several cars models ...
0
votes
2answers
14 views
How to reduce the Root mean square error
I have dataset which describe "how many passenger arriving in some airport " and I would like to predict how many passenger arriving in monthly bases for next year. The features that I have is the ...
1
vote
3answers
46 views
Correlation between categorical variables based on the target distribution
Let $X$ be a category with very high cardinality and $Y$ be my target. when I look at $X$ distribution to $Y$ I see that some of the levels are very similar to each other . I would like to find a way ...
7
votes
1answer
141 views
Why does frequency encoding work?
Frequency encoding is a widely used technique in Kaggle competitions, and many times proves to be a very reasonable way of dealing with categorical features with high cardinality. I really don't ...
2
votes
3answers
41 views
Aggregate categorical feature by the target
Having a list of triplets {X1,X2,Y} such as :
{pennsylvania, fever , malaria}
{pennsylvania, headache , malaria}
{arizona, ketone smell , flu}
{new york, fever , cancer}
{ohio, hand pain , ...
1
vote
0answers
12 views
Creating Flags Instead of Designated Values
I'm working with http://archive.ics.uci.edu/ml/datasets/Bank+Marketing# dataset in order to create a model. We're going to use it in a presentation to introduce people our new data science environment....
3
votes
2answers
44 views
Should features be correlated on uncorrelated for classification and regression (prediction)
I have seen researchers using pearson's correlation coefficient to find out the relevant features -- to keep the features that have a high correlation value with the target. The physical implication ...
1
vote
2answers
32 views
1
vote
0answers
14 views
Can feature representation acquired by a same model but trained on different corpus be used on the same classification model?
For example, if I wanna do document classification with doc2vec embeddings, first I train the training set to get doc2vec embeddings, and fit the embeddings to a classification model; later on when I ...
0
votes
0answers
8 views
LIME- how does sampling of data happen at observation( data point) of interest?
I was going through paper of LIME( Locally interpret-able model agnostic explanation)
I am wondering how exactly data is sampled at point of interest to fit regression?
https://arxiv.org/pdf/1602....
1
vote
2answers
27 views
Does the predict function in machine learning understand categorical data
I understand that before feature engineering one has to split the dataset into train and test data, so as to avoid bias in the analysis. I also understand that the machine learning model does not ...
1
vote
0answers
22 views
Feature Vectors representation
I would like to know I how you represent a feature vector like this dataset wise. The vector length is dynamic but the each element has a fixed length (9). For xgboost implementation, do I just create ...
0
votes
0answers
25 views
how to use sklearn without feature selection
I am trying to study the effect of using feature selection onmy text classification code . I want to make a rating without any feature selection, but sklearn use document frequency (df) by default
...
0
votes
0answers
15 views
Can I merge levels or factors having equal mean in categorical variable
I compared levels of categorical variable by their respected mean, obtain from continuous response variable using pivot table. I found that some of the levels is having nearly equal mean e.g 'BrDale' ...
1
vote
0answers
23 views
Extracting Features for Graph transformation
Suppose I have a directed graph G (V,E) whose transformation is defined by a library of patterns. Each vertex is of particular type.
The library of patterns contain subgraphs (g1,g2,g3 etc)and it's ...
0
votes
0answers
19 views
Feature engineering ideas with dates, coordinates and other variables
I'm working on an ETA problem where I'm trying to estimate a time of arrival for a delivery. I have coordinates of pickup/destination, time of pick , infos about the rider, some other variables that ...
0
votes
1answer
30 views
Feature engineering - house price prediction (small dataset) [closed]
I am working on the task of predicting real estate prices. My dataset has only 10 variables described below. I'm thinking about feature engineering but nothing comes to mind.
Variables:
...
1
vote
1answer
19 views
Why does removal of some features improve the performance of random forests on some occasions?
I completed feature importance of a random forest model. I removed the bottom 4 features out of 17 features. The model performance actually improved. Shouldn't the performance degrade after removal of ...
0
votes
0answers
8 views
Accelerometer and Gyroscope features
I am having accelerometer and gyroscope reading along x,y,z axis and want to get motion direction info at each time step. What all feature extraction would be best suited for this type of requirement. ...
0
votes
0answers
14 views
Optimal practices to group data by Customer ID for churn prediction
Here's a quite common problem and I read a couple of questions/answers on it, however I still having my doubts about what are the best practices for grouping data by Customer ID for churn prediction.
...
-1
votes
1answer
18 views
How to interact two variables in python?
I have about 8 features as my predictors in a logistic model I am trying to fit in python. One of the features is TotalAward (Financial Aid) and the second is NEED. I am attempting to predict the ...
1
vote
0answers
13 views
Variable importance of Numerical features in Classification Model - Random Forest Classifier
I have few numeric features in my model. Out of 25 features, I have 7-8 numeric features in my model.
One thing I observed is model gives more weightage to numerical feature compare to categorical ...
0
votes
1answer
31 views
How to Avoid rarely used discrete feature values in a dataset
On Google's ML crash course it states:
Good feature values should appear more than 5 or so times in a data
set. Doing so enables a model to learn how this feature value relates
to the label. ...
0
votes
1answer
32 views
Aggregate Categorical Data
I have a scenario in which I'm required to run my analysis at the Account level. One of the features that I'd like to look at is the no. of subscriptions against an account. There can be multiple ...
2
votes
2answers
50 views
Dealing with NaN for predictive models
I have data set that has data for patients:
Arrival_Date : is when the patient has arrived
Seen_By_Nurse : is number of minutes patient take to be seen by nurse since arrival
when value is NaN it ...
2
votes
1answer
61 views
Feature selection is not that useful?
I've been doing a few DataScience competitions now, and i'm noticing something quite odd and frustrating to me. Why is it frustrating? Because , in theory, when you read about datascience it's all ...
0
votes
1answer
19 views
Adding time as a feature with xgboost/random forests
I am trying to use xgboost for performing some regression and the features I have are rather simple and limited. I have the time stamp associated with some ...
0
votes
0answers
13 views
Should I convert noncontinuous numerical values to categorical features?
I'm working with very sparse matrixes and have several noncontinuous numerical fields. Are these values better utilized as continuous (prevent from increasing sparseness) or as categorical features?
...
0
votes
0answers
20 views
Deep learning, signal processing and feature engineering
I have a signals represented in python in dense matrices (the values are y-coordinates from a chart - eg. weather temp etc. in different locations around the world).
I'm currently trying to process/...
1
vote
0answers
19 views
How can I train a machine learning model with below characterstics? [closed]
Hi I have a classifier model to solve, which has close to 56k samples and 30 features which ...
0
votes
1answer
90 views
xgboost feature selection and feature importance
when you create the new feature for data analysis
for linear regression, it is clear that the feature has to be linear with other features is better
but for xgboost what is the guideline to make a ...
0
votes
0answers
51 views
Mean encoding for linear regression: leveraging domain expertise
I'm trying to build a linear model to predict a customer satisfaction score that measures the overall store experience. My customer could interact with my store using an offline channel (physical) or ...
1
vote
1answer
20 views
Filling created feature with values
I'm trying to improve accuracy. I created a few new features based on old features. So I need to fill new feature's empty cell with same values in order to equaling shapes.Then, I tried it with median ...
2
votes
0answers
78 views
Can you do automated feature engineering in R?
Since now Python has its own deep feature synthesis library, is there anything similar available in R? I know of bounceR but I'm not sure if it does the job as DFS does. Is anyone aware of anything ...
0
votes
1answer
26 views
What should be the criteria to select features using correlation factors between features? [duplicate]
For the titanic dataset, I have done some feature engineering (one-hot encoded the features) and now I have developed a heatmap to view the correlation between different features.
I'm not able to ...
1
vote
1answer
55 views
To One-Hot-Encode or not to One-Hot-Encode?
I have been struggling to find proof for that but I couldnt
Every time I prepare dataset I face the same issue
when a column is a classification such as ...
0
votes
0answers
20 views
Time Series Regression and Exponentially weighted mean of lag values-For Advanced Time Series Experts
I need help for a time series regression problem in engineering features.
Background: The dataset has weekly data for sales/orders of hundreds of products for last 145 weeks totalling 450000 ...
0
votes
0answers
24 views
Feature selection for circular data in time-series
I'm predicting ozone concentration based on meteorological and seasonal variables. In the feature engineering stage I converted the MONTH, DAY_OF_WEEK, DAY_OF_YEAR to its sin and cosine components ...
0
votes
0answers
14 views
What are the common feature extraction technic to compare 2 sequence of timestamps?
I am building some predictive models for an online shopping site. I have timestamp log of customers' arrival time, time spent on a product's webpage, purchased or not, and a few others.
I have tried ...
0
votes
1answer
17 views
How to include both origin and destination in your features?
I'm trying to predict the price of transportation for trucking freight. Two important features that I think would be of great impact are Origin and Destination. What's the best way to include that in ...
1
vote
1answer
27 views
For a regression model, can you transform all your features to linear to make a better prediction?
I was thinking. Would it be a good approach to check your features one by one (assuming you have a manageable amount of them) and see the relationship they have with your target variable, if they ...
0
votes
1answer
34 views
Feature engineering for time series (audio signals)
I have a task to perform classification of audio signals using any suitable algorimth. After some research I found out, that CNN from this paper shows promising results. However, it still needs to be ...
0
votes
0answers
11 views
Can we vectorize nominal text feature using tfidf or count vectorizer?
I recently participated in the hackathon. The dataset includes drug name, sentiments about the drug, unique id. The target variable is the sentiment.
It was a sentiment classification or analysis ...
2
votes
1answer
49 views
Blind feature engineering
I received a dataset for analysis that had ~100 numeric columns with anonymous column names(X1, X2, X3, etc...) and asked to do a binary classification. My resulting classification algorithm using a ...
0
votes
0answers
31 views
How to time series forecast with multiple time series data sets on the same time series index
How does time series work with multiple time series data sets on the same index?
For example, suppose I were a utilities company. Suppose I have the electricity usage of two homes, each indexed for ...
2
votes
2answers
42 views
Training vs test data set for supervised learning in real life scenario
In the tutorials, I have noticed only similar data has been used with models training and prediction.
I was wondering how cases where you can't find training data that is similar to your final use ...
0
votes
3answers
49 views
Can we use median to replace all the missing values from a column
If we can use the median to replace all of the missing values from a column, then what is the advantage of using median over mean for replacing the missing values in a column with numeric values?
0
votes
1answer
18 views
Using strong predictor in Model training?
I am trying to build a Disease predictor based on symptoms. I am using data scraped from Symcat website. After sampling the data we have symptoms to disease mapped for training purpose.Data looks like ...
0
votes
0answers
9 views
What kind of feature engineering suits for random generation like time series data
My time series data looks like random generation which was obtained by aggregation day-wise sales of a product. The dependent variable does not show any kind of pattern.
So have tried tries different ...
