Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Paper: An intelligent shopping list #465

Open
wants to merge 102 commits into
base: 2019
from

Conversation

Projects
None yet
5 participants
@TahiriNadia
Copy link

commented May 21, 2019

If you are creating this PR in order to submit a draft of your paper,
see http://procbuild.scipy.org/ for logs generated by the build
process.

See the project readme
for more information.

TahiriNadia and others added some commits May 10, 2019

Edited Intro + Proposal
Use of gender neutral term "they/their" instead of masculine "he/his". Multiple grammatical consistency improvements + changing names
Edited Datasets
Sentence structures

TahiriNadia added some commits Jun 20, 2019

@TahiriNadia

This comment has been minimized.

Copy link
Author

commented Jun 22, 2019

Thank you @TahiriNadia for this interesting paper! I have reviewed it and I think it is mostly in good shape. I have made many comments inline, many of which are wording or grammar suggestions; some others are more substantial.

In particular, I suggest to move Figure 5 (graphical illustration) to the start of the model section, and to relate the NNMF piece more clearly to the text describing the model -- preferably with a NNMF subsection.

I look forward to iterating with you on the comments.

Thank you for your comments. Following your recommendations, we moved Figure 5 to the start of the model section and related the NNMF method to the text describing the model.

@TahiriNadia

This comment has been minimized.

Copy link
Author

commented Jun 22, 2019

Great paper! I look forward to having your algorithms organize my shopping lists 😄 . I think I may have missed the part where you talk about predicting which store people will purchase from -- it's a bit past my bedtime right now. Was this somewhere in the methods?

Thank you. In this first version of the algorithm, the prediction of the store people will purchase from was not yet performed, but on the perspective, we plan to investigate this question.

@TahiriNadia

This comment has been minimized.

Copy link
Author

commented Jun 23, 2019

@deniederhut @cbcunc and @stargaser I believe it should be ready for a second review :-)

@deniederhut
Copy link
Member

left a comment

I like the new tables! I think they add a lot of useful detail about the work that you have done, and about the data that went into the model.

I'm still a little unclear about the overall architecture of your pipeline. Where does the total basket size fit in the prediction? Where does the prediction about product categories go? What about the user clustering?

I think this might be a specific case of something more general that I noticed. There are a lot of great details in the paper, and it would be easier to put them in context if they were organized together. Maybe think about organizing the methods to have one section for each experiment you did?

Show resolved Hide resolved papers/nadia_tahiri/nadia_tahiri.rst Outdated
The IDs (`user\_id`) are not sequentially. In total, we have 1374 users. Among them we have 374 real users and 1000 users whose behaviour was generated following the distribution of real users (see Figure 3) and
the consumer statistics available in the report by Statistics Canada (2017).

The product categories were determined with the purchase histories is the current version of our model

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

I'm having trouble parsing this sentence. Was it meant to read "in the current version of our model, which does not allow"?

Show resolved Hide resolved papers/nadia_tahiri/nadia_tahiri.rst Outdated
================

We found that there are 3 consumer profiles see [WJ03]_, [WJ02]_, and [TNTK16]_.
These values were obtained from Statistics Canada. Moreover, the distributions used our study follow these statistics.

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

How was this determined? Tweaking the k in k means? Chi-square on expected values?

Show resolved Hide resolved papers/nadia_tahiri/nadia_tahiri.rst Outdated

.. math::
\max_\mathcal{P} \mathbb{E}_{p'\in \mathcal{P}}[F_1(\mathcal{P})]=\max_\mathcal{P}\mathbb{E}_{p'\in \mathcal{P}}\bigg[\frac{2\sum_{i\in \mathcal{P}}\text{TP}(i)}{\sum_{i\in \mathcal{P}}(2\text{VP}(i)+\text{FN}(i)+\text{FP}(i))}\bigg],

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

I don't think it is necessary to define all of these metrics.

We present the obtained results using proposed method in this section.
As well as the metrics (see Equations 1-5) that are utilized to evaluate the performance of methods.

Statistic score

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

I'm not sure we need this section here either. Maybe pare it down to just the description of f1, since that is what you are using for thresholding?

Results
-------

Figure :ref:`productpca` illustrates PCA of 20 random products projected into 2 dimensions.

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

Did your model use PCA to generate clusers of products? If so, where does it fit in your model architecture? If not, what point are you making by including it in the paper?

:scale: 21%
:figclass: wt

Distribution of :math:`F_1` measures against rebates (a) and stores (b). :label:`violon`

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

What am I supposed to be learning about your model from this plot?

\hline
\textbf{Product} & \textbf{Number of baskets} \\
\hline
Banana & 6138 \\

This comment has been minimized.

Copy link
@deniederhut

deniederhut Jun 25, 2019

Member

I find this table very surprising. I would have expected to see things like bread and potato chips in the top ten products. Do people really eat more lemons, limes, avocados, and bananas than bread?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.