Blog

After struggling with the report for a few weeks, the first draft is finally done. It is at least half in point form, but all the necessary ideas are in there. The draft is already sent to Dr Kwan for comments. Now I can put more time facilitating YouFind in porting the model for their use of their website.

In the last two weeks I have been writing the draft of the report.

One thing that is left to do with YouFind apart from delivering the product is the evaluation of the model with them. So today I discussed with Vincent an objective way to evaluate the machine learning model against the their initial linear model. We have agreed that we will use Spearman's Ranking Coefficient as the metric to compare the models.

Apart from this discussion, Vincent also told me that they want to use the new model in their website content for attracting clients. This use is on top of what has been discussed, but I am more than happy to facilitate them as this proves their satisfaction of my work.

Today I met with the people of the sponsor at their premises. I showed Vincent and Raymond the progress of the project and the preliminary results of the machine learning algorithms. We agreed to filter out certain low quality data to see if the results can be improved. We also touched on a little bit how to improve the accuracy of the model in the long run and how the process will be when the model is put in use.

I only set up the meeting with Academic Supervisor Dr. Peter Kwan yesterday and he proposed to meet today early this morning. I want to thank him for his enthusiasm. With that, I don't mind crawling out of bed early in the morning and travel half an hour to the meeting place.

I told him about the r^2 which disappoints me a bit. He encouraged me that the r^2 is not the most important thing in the project. It's the process and argument / logical thinking that matter. I am slightly relieved. However, I do really want to do something great to the sponsor that had spent many hours with me in the project. So I do really hope that despite the relatively low r^2, I can still produce a tool that is useful to YouFind.

It takes a couple of hours to get the data ready for machine learning. One task is to identify relevant characteristics of the search terms that needs to be added to the dataset. This requires me to see the terms that have the highest and lowest average ratings and identify any discernible common characteristics in some of them and write script to create such features in the dataset. This take a lot of time and concentration because I could only do that by eyeballing the search terms!

After this round, I don't get very high r^2 from the models. However, r^2 is not the measurement. High r^2 can boost my confidence but what really matters is whether YouFind can rely on the list generated by the algorithms and depend on it to increase their productivity. I look forward to that part.

For now, I have to talk to my supervisor to ensure I am doing the right thing. I'll meet him tomorrow.

My Planning Review has been completed in mid-December last year. Both the sponsor and the Academic Supervisor have signed the respective part of the TOR at that time. Because UOS never gave me a revised deadline for the Planning Review, I have not uploaded the Planning Review to Canvas, until today. However, Academic Supervisor Dr Peter Kwan has already received the Planning Review informally in December, and we have met several times since then.

I asked the Hong Kong office several times about the deadline. Since it was never set, the office did not force me to formally submit the document. Finally, both I and the office are tired of the conversation :-), so we agreed that I submitted today. The only thing I added to the Planning Review is a paragraph about how the pandemic is affecting the data collection process. Unfortunately, we are now 4 weeks behind schedule! I do hope that eventually the deadline given by UOS in the meeting later in the month will give me enough time to finish the project nicely!

Dear SEO Team,

As of the Tuesday deadline all of you have finished all the questionnaires.  Thank you very much for your support! 

It is now time for me to dig into the data and hopefully produce a machine learning model that can identify the high value search terms to help you in your work.  I will come back to you later for the results.

Again, thank you!

Victor

As of the evening of March 10, all questionnaires are complete!

The part of the project has been an extremely long journey. It was originally schedule to be finished within a month. Now it has taken more than 2 months directly and indirectly because of the work arrangement of YouFind as the reaction to the spread of the new corona virus.

The development of the machine learning starts now! Hope the quality of the data is good enough and support the machine learning process well!

Dear SEO Team,

I checked the questionnaires completed and I found that only a few of you have completed the batch that was due last Friday. I understand that the number of questionnaires released last week is a but higher than the previous weeks. In any case, please complete the batch as soon as possible.

The final batch is originally scheduled to be due this Friday. Looking at the progress I think it makes sense to give you a bit more time so that we will have better response quality. I am extending the deadline to next Tuesday (March 10) for the final batch.

If you run into any problems, please let me know.

Thank you very much!

Dear SEO Team,

Most of you have already completed the first 5 questionnaires of Batch 2, and many of you have also done part of Batch 3. Thank you very much!

As mentioned last week, please finish the rest of Batch 2 and the whole Batch 3 by Friday end of day (February 28).

Batch 4, the final batch, is now released. Please complete the batch by next week on Friday (March 6). By then this stage will be complete.

If you run into any problems, please let me know.

Thank you very much!

Dear SEO Team,

I notice that some of you still have not started with Batch 2. If you are one of them, please start now and finish the 5 questionnaires (first 5 of Batch 2) by end of tomorrow.

The rest of Batch 2 and Batch 3 are now open. Please finish them by end of next week. The final batch, Batch 4, will be open next week.

As always, if you encounter any problems, please let me know.

Thank you very much!

Today I visited YouFind again to talk with the SEO Team members who were not in the kick-off meeting. The purpose of the meeting is to ensure that they understand the project so that they would be motivated to provide the best input to the project. I was pleasantly surprised that they were enthusiastic. This makes me more confident that the input collected from the team will be of high quality.