I would like to use a multinomial logistic regression to get win probabilities for each of the 5 horses that participate in any given race using each horses previous average speed.
RACE_ID H1_SPEED H2_SPEED H3_SPEED H4_SPEED H5_SPEED WINNING_HORSE
1 40.482081 44.199627 42.034929 39.004813 43.830139 5
2 39.482081 42.199627 41.034929 41.004813 40.830139 4
I am stuck on how to handle the independent variables for each horse given that any of the 5 horses average speed can be placed in any of H1_SPEED through H5_SPEED.
Given the fact that for each race I can put any of the 5 horses under H1_SPEED meaning there is no real relationship between H1_SPEED from RACE_ID 1 and H1_SPEED from RACE_ID 2 other than the arbitrary position I selected.
Would there be any difference if the dataset looked like this -
For RACE_ID 1 I swapped H3_SPEED and H5_SPEED and changed WINNING_HORSE from 5 to 3
For RACE_ID 2 I swapped H4_SPEED and H1_SPEED and changed WINNING_HORSE from 4 to 1
RACE_ID H1_SPEED H2_SPEED H3_SPEED H4_SPEED H5_SPEED WINNING_HORSE
1 40.482081 44.199627 43.830139 39.004813 42.034929 3
2 41.004813 42.199627 41.034929 39.482081 40.830139 1
Is this an issue, if so how should this be handled? What if I wanted to add more independent features per horse?
You cannot change in that way your dataset, because each feature (column) has a meaning and probably it depends on the values of the other features. You can imagine it as a six dimensional hyperplane, if you change the value of a feature the position of the point in the hyperplane changes, it does not remain stationary.
If you deem that a feature is useless to solve your problem (i.e. it is independent from the target), you can drop it or avoid to use it during the training phase of your model.
Edit
To solve your specific problem you may add a parameter for each speed column that takes care of the specific horse which is running with that speed. It is a sort of data augmentation, in order to add more problem related features to your model.
RACE_ID H1_SPEED H1_HORSE H2_SPEED H2_HORSE ... WINNING_HORSE
1 40.482081 1 44.199627 2 ... 5
2 39.482081 3 42.199627 5 ... 4
I've invented the number associated to each horse, but it seems that this information is present in your dataset.
I'm trying to display weighted random results from my database and I'm unable to get results with expected accuracy. I've followed what I learnt here and here.
This would be my table:
+--------+-----------+
| weight | image |
+--------+-----------+
| 50 | A |
| 25 | B |
| 25 | C |
+--------+-----------+
I need the image A to appear 50% of the times, the image B the other 25% of the times and C the remaining 25% of the times.
The SQL estatement I'm using goes like this:
SELECT image FROM images WHERE weight > 0 ORDER BY -LOG(1.0 - RAND()) / weight LIMIT 10
So in order to test this properly I made a php script to have this iterate 10,000 times, counting how many times a, b or c was being shown and I display the results on my test script with percentages, like this:
a total: 4976 - 49,76%
b total: 2538 - 25,38%
c total: 2486 - 24,86%
With only 10,000 results and considering the RAND() is just a randomization function I would consider this results to be accurate enough. The problem is that I run this script about 100 times and I realized that 98 out of 100 times b had a higher percentage count than c.
I'm trying to understand what's wrong, both values (b and c) on the table are the same and I'm not introducing any other ordering factor. I took it up a notch and I went for 100,000 iterations of the SQL clause. These are the results:
a total: 50185 - 50,185%
b total: 25201 - 25,201%
c total: 24614 - 24,614%
I run this last test about 50 times (with long wait times between each). This time b was above c every time and accuracy was worse than the accuracy at 10000 iterations. You would expect that as you go higher on the number of iterations, the percentage variation should be getting smaller and the results more accurate. It's obvious that either I'm doing something wrong or RAND() is not really random enough.
Matematically speaking if it was perfectly random it should be improving accuracy the more iterations you make and not the opposite.
Any explanation/solution is welcome.
I'm currently ranking videos on a website using a bayesian ranking algorithm, each video has:
likes
dislikes
views
upload_date
Anyone can like or dislike a video, a video is always views + 1 when viewed and all videos have a unique upload_date.
Data Structure
The data is in the following format:
| id | title | likes | dislikes | views | upload_date |
|------|-----------|---------|------------|---------|---------------|
| 1 | Funny Cat | 9 | 2 | 18 | 2014-04-01 |
| 2 | Silly Dog | 9 | 2 | 500 | 2014-04-06 |
| 3 | Epic Fail | 100 | 0 | 200 | 2014-04-07 |
| 4 | Duck Song | 0 | 10000 | 10000 | 2014-04-08 |
| 5 | Trololool | 25 | 30 | 5000 | 2014-04-09 |
Current Weighted Ranking
The following weighted ratio algorithm is used to rank and sort the videos so that the best rated are shown first.
This algorithm takes into account the bayesian average to give a better overall ranking.
Weighted Rating (WR) = ((AV * AR) + (V * R))) / (AV + V)
AV = Average number of total votes
AR = Average rating
V = This items number of combined (likes + dislikes)
R = This items current rating (likes - dislikes)
Example current MySQL Query
SELECT id, title, (((avg_vote * avg_rating) + ((likes + dislikes) * (likes / dislikes)) ) / (avg_vote + (likes + dislikes))) AS score
FROM video
INNER JOIN (SELECT ((SUM(likes) + SUM(dislikes)) / COUNT(id)) AS avg_vote FROM video) AS t1
INNER JOIN (SELECT ((SUM(likes) - SUM(dislikes)) / COUNT(id)) AS avg_rating FROM video) AS t2
ORDER BY score DESC
LIMIT 10
Note: views and upload_date are not factored in.
The Issue
The ranking currently works well but it seems we are not making full use of all the data at our disposal.
Having likes, dislikes, views and an upload_date but only using two seems a waste because the views and upload_date are not factored in to account how much weight each like / dislike should have.
For example in the Data Structure table above, items 1 and 2 both have the same amount of likes / dislikes however item 2 was uploaded more recently so it's average daily views are higher.
Since item 2 has more likes and dislikes in a shorter time than those likes / dislikes should surely be weighted stronger?
New Algorithm Result
Ideally the new algorithm with views and upload_date factored in would sort the data into the following result:
Note: avg_views would equal (views / days_since_upload)
| id | title | likes | dislikes | views | upload_date | avg_views |
|------|-----------|---------|------------|---------|---------------|-------------|
| 3 | Epic Fail | 100 | 0 | 200 | 2014-04-07 | 67 |
| 2 | Silly Dog | 9 | 2 | 500 | 2014-04-06 | 125 |
| 1 | Funny Cat | 9 | 2 | 18 | 2014-04-01 | 2 |
| 5 | Trololool | 25 | 30 | 5000 | 2014-04-09 | 5000 |
| 4 | Duck Song | 0 | 10000 | 10000 | 2014-04-08 | 5000 |
The above is a simple representation, with more data it gets a lot more complex.
The question
So to summarise, my question is how can I factor views and upload_date into my current ranking algorithm in a style to improve the way that videos are ranked?
I think the above example by calculating the avg_views is a good way to go but where should I then add that into the ranking algorithm that I have?
It's possible that better ranking algorithms may exist, if this is the case then please provide an example of a different algorithm that I could use and state the benefits of using it.
Taking a straight percentage of views doesn't give an accurate representation of the item's popularity, either. Although 9 likes out of 18 is "stronger" than 9 likes out of 500, the fact that one video got 500 views and the other got only 18 is a much stronger indication of the video's popularity.
A video that gets a lot of views usually means that it's very popular across a wide range of viewers. That it only gets a small percentage of likes or dislikes is usually a secondary consideration. A video that gets a small number of views and a large number of likes is usually an indication of a video that's very narrowly targeted.
If you want to incorporate views in the equation, I would suggest multiplying the Bayesian average you get from the likes and dislikes by the logarithm of the number of views. That should sort things out pretty well.
Unless you want to go with multi-factor ranking, where likes, dislikes, and views are each counted separately and given individual weights. The math is more involved and it takes some tweaking, but it tends to give better results. Consider, for example, that people will often "like" a video that they find mildly amusing, but they'll only "dislike" if they find it objectionable. A dislike is a much stronger indication than a like.
I can point you to a non-parametric way to get the best ordering with respect to a weighted linear scoring system without knowing exactly what weights you want to use (just constraints on the weights). First though, note that average daily views might be misleading because movies are probably downloaded less in later years. So the first thing I would do is fit a polynomial model (degree 10 should be good enough) that predicts total number of views as a function of how many days the movie has been available. Then, once you have your fit, then for each date you get predicted total number of views, which is what you divide by to get "relative average number of views" which is a multiplier indicator which tells you how many times more likely (or less likely) the movie is to be watched compared to what you expect on average given the data. So 2 would mean the movie is watched twice as much, and 1/2 would mean the movie is watched half as much. If you want 2 and 1/2 to be "negatives" of each other which sort of makes sense from a scoring perspective, then take the log of the multiplier to get the score.
Now, there are several quantities you can compute to include in an overall score, like the (log) "relative average number of views" I mentioned above, and (likes/total views) and (dislikes / total views). US News and World Report ranks universities each year, and they just use a weighted sum of 7 different category scores to get an overall score for each university that they rank by. So using a weighted linear combination of category scores is definitely not a bad way to go. (Noting that you may want to do something like a log transform on some categories before taking the linear combination of scores). The problem is you might not know exactly what weights to use to give the "most desirable" ranking. The first thing to note is that if you want the weights on the same scale, then you should normalize each category score so that it has standard deviation equal to 1 across all movies. Then, e.g., if you use equal weights, then each category is truly weighted equally. So then the question is what kinds of weights you want to use. Clearly the weights for relative number of views and proportion of likes should be positive, and the weight for proportion of dislikes should be negative, so multiply the dislike score by -1 and then you can assume all weights are positive. If you believe each category should contribute at least 20%, then you get that each weight is at least 0.2 times the sum of weights. If you believe that dislikes are more important that likes, then you can say (dislike weight) >= c*(like weight) for some c > 1, or (dislike_weight) >= c*(sum of weights) + (like weight) for some c > 0. Similarly you can define other linear constraints on the weights that reflect your beliefs about what the weights should be, without picking exact values for the weights.
Now here comes the fun part, which is the main thrust of my post. If you have linear inequality constraints on the weights, all of the form that a linear combination of the weights is greater than or equal to 0, but you don't know what weights to use, then you can simply compute all possible top-10 or top-20 rankings of movies that you can get for any choice of weights that satisfy your constraints, and then choose the top-k ordering which is supported by the largest VOLUME of weights, where the volume of weights is the solid angle of the polyhedral cone of weights which results in the particular top-k ordering. Then, once you've chosen the "most supported" top-k ranking, you can restrict the scoring parameters to be in the cone that gives you that ranking, and remove the top k movies, and compute all possibilities for the next top-10 or top-20 ranking of the remaining movies when the weights are restricted to respect the original top-k movies' ranking. Computing all obtainale top-k rankings of movies for restricted weights can be done much, much faster than enumerating all n(n-1)...(n-k+1) top-k possible rankings and trying them all out. If you have two or three categories then using polytope construction methods the obtainable top-k rankings can be computed in linear time in terms of the output size, i.e. the number of obtainable top-k rankings. The polyhedral computation approach also gives the inequalities that define the cone of scoring weights that give each top-k ranking, also in linear time if you have two or three categories. Then to get the volume of weights that give each ranking, you triangulate the cone and intersect with the unit sphere and compute the areas of the spherical triangles that you get. (Again linear complexity if the number of categories is 2 or 3). Furthermore, if you scale your categories to be in a range like [0,50] and round to the nearest integer, then you can prove that the number of obtainable top-k rankings is actually quite small if the number of categories is like 5 or less. (Even if you have a lot of movies and k is high). And when you fix the ordering for the current top group of movies and restrict the parameters to be in the cone that yields the fixed top ordering, this will further restrict the output size for the obtainable next best top-k movies. The output size does depend (polynomially) on k which is why I recommended setting k=10 or 20 and computing top-k movies and choosing the best (largest volume) ordering and fixing it, and then computing the next best top-k movies that respect the ordering of the original top-k etc.
Anyway if this approach sounds appealing to you (iteratively finding successive choices of top-k rankings that are supported by the largest volume of weights that satisfy your weight constraints), let me know and I can produce and post a write-up on the polyhedral computations needed as well as a link to software that will allow you to do it with minimal extra coding on your part. In the meantime here is a paper http://arxiv.org/abs/0805.1026 I wrote on a similar study of 7-category university ranking data where the weights were simply restricted to all be non-negative (generalizing to arbitrary linear constraints on weights is straightforward).
A simple approach would be to come up with a suitable scale factor for each average - and then sum the "weights". The difficult part would be tweaking the scale factors to produce the desired ordering.
From your example data, a starting point might be something like:
Weighted Rating = (AV * (1 / 50)) + (AL * 3) - (AD * 6)
Key & Explanation
AV = Average views per day:
5000 is high so divide by 50 to bring the weight down to 100 in this case.
AL = Average likes per day:
100 in 3 days = 33.33 is high so multiply by 3 to bring the weight up to 100 in this case.
AD = Average dislikes per day:
10,000 seems an extreme value here - would agree with Jim Mischel's point that dislikes may be more significant than likes so am initially going with a negative scale factor of twice the size of the "likes" scale factor.
This gives the following results (see SQL Fiddle Demo):
ID TITLE SCORE
-----------------------------
3 Epic Fail 60.8
2 Silly Dog 4.166866
1 Funny Cat 1.396528
5 Trololool -1.666766
4 Duck Song -14950
[Am deliberately keeping this simple to present the idea of a starting point - but with real data you might find linear scaling isn't sufficient - in which case you could consider bandings or logarithmic scaling.]
Every video have:
likes
dislikes
views
upload_date
So we can deduct the following parameters from them:
like_rate = likes/views
dislike_rate = likes/views
view_rate = views/number_of_website_users
video_age = count_days(upload_date, today)
avg_views = views/upload_age
avg_likes = likes/upload_age
avg_dislikes = dislikes/upload_age
Before we can set the formula to be used, we need to specify how different videos popularity should work like, one way is to explain in points the property of a popular video:
A popular video is a recent one in most cases
The older a video gets, the higher avg_views it requires to become popular
A video with a like_rate over like_rate_threshold or a dislike_rate over dislike_rate_threshold, can compete by the difference from its threshold with how old it gets
A high view_rate of a video is a good indicator to suggest that video to a user who have not watched it before
If avg_likes or avg_dislikes make most of avg_views, the video is considered active in the meantime, in case of active videos we don't really need to check how old it's
Conclusion: I don't have a formula, but one can be constructed by converting one unit into another's axis, like cutting a video age by days based on a calculation made using avg_likes, avg_dislikes, and avg_views
Since no one has pointed it out yet (and I'm a bit surprised), I'll do it. The problem with any ranking algorithm we might come up with is that it's based on our point of view. What you're certainly looking for is an algorithm that accomodates the median user point of view.
This is no new idea. Netflix had it some time ago, only they personalized it, basing theirs on individual selections. We are looking - as I said - for the median user best ranking.
So how to achieve it? As others have suggested, you are looking for a function R(L,D,V,U) that returns a real number for the sort key. R() is likely to be quite non-linear.
This is a classical machine learning problem. The "training data" consists of user selections. When a user selects a movie, it's a statement about the goodness of the ranking: selecting a high-ranked one is a vote of confidence. A low-ranked selection is a rebuke. Function R() should revise itself accordingly. Initially, the current ranking system can be used to train the system to mirror its selections. From there it will adapt to user feedback.
There are several schemes and a huge research literature on machine learning for problems like this: regression modeling, neural networks, representation learning, etc. See for example the Wikipedia page for some pointers.
I could suggest some schemes, but won't unless there is interest in this approach. Say "yes" in comments if this is true.
Implementation will be non-trivial - certainly more than just tweaking your SELECT statement. But on the plus side you'll be able to claim your customers are getting what they're asking for in very good conscience!
Hypothetical database for events happening around the world.
EVENT
event_id | event_name
1 | Great Wall Party
2 | Times Square Dance
3 | Sydney Blowout
PLACE
place_id | place_name
54 | Times Square
55 | Manhattan
56 | New York City
57 | New York State
58 | USA
EVENTPLACE
eventid | placeid
2 | 54
RELATEDPLACES
rel_placeid1 | rel_placeid2
54 | 55
55 | 56
56 | 57
57 | 58
If I display the event, Times Square Dance, I’d like to display all the places that appear up the chain of its associated places via the RELATEDPLACES table (i.e. Times Square, Manhattan, New York City, New York State, USA). Likewise, if I call all events for USA, I’d like the Times Square dance to be listed, given its EVENTPLACE (Times Square) appears at the bottom of the RELATEDPLACES chain of associations starting with USA.
I think I need to create an inner loop within my SQL command so that it keeps performing until there is a break in the chain. So far (using the first of the two above examples) I have:-
SELECT place_nm FROM eventplace
INNER JOIN relatedplaces ON placeid = rel_placeid1
INNER JOIN place ON rel_placeid2 = place_id
[where the loop should begin:
INNER JOIN relatedplaces ON place_id = rel_placeid1
INNER JOIN place ON rel_placeid2 = place_id
end loop]
WHERE eventid = ‘2’;
This is complicated by the fact that I need different table aliases for each loop, which means I can’t state in the opening SELECT statement that I want to be collecting all the place_name data in the same column.
I’m not sure if I what I am trying to achieve is even possible and my current fallback solution is to list all of Times Square’s related places in the RELATEDPLACES table, rather than just the next largest place (Manhattan), but this seemed like the better solution (and would also save database space).
Can anyone suggest the SQL SELECT command I might need to use? Cheers!
Quite an intriguing problem. But the issue is that MySQL currently does not allow for recursive queries. You have two options :
Decided on the level of RelatedEvents upto which you want to dig into, and create a query for the same (using excel or a small C code)
Use a program to generate and execute the queries on the fly; recursively make DB queries from the program. I suggest this is the best option, though requires repeated DB access.
I want to calculate distance between two zip codes before inserting the data to database .basically i have a these tables
zip code table
| zipcode | lat | long |
01230 60.1756 23.12
01240 60.1756 25.25
customer table
| name | zip code |
foo 01230
sales man table
| name | zip code | workingdistanceinkm
foo 01240 200
foo1 01230 100
What I want to do is calculate the distance between the sales mans and the customer if it is any of the salesman working area before the data of the customer is inserted to customer table .
MY approach was to calculate the distance between a customer and every sales man that is in the sales man table. But this makes a lot of queries for example if I have 1000 sales man it means I am calculating the distance between the new customer data to be inserted with those every one of the 1000 salesmen.
I am wondering if it's possible to write one query to do the same task.
Have a look at
www.zipcodeworld.com/samples/distance.php.html
note that distance calculations between zipcodes are not always the actual representations of the distance. This is just the distance of an imaginary straight line between the two points. But in reality it is longer
Below URL helped me a lot. Please check "Finding Locations with MySQL" section. Thanks.
https://developers.google.com/maps/articles/phpsqlsearch_v3