Time Prediction based on existing date:time records - php

I have a system that logs date:time and it returns results such as:
05.28.2013 11:58pm
05.27.2013 10:20pm
05.26.2013 09:47pm
05.25.2013 07:30pm
05.24.2013 06:24pm
05.23.2013 05:36pm
What I would like to be able to do is have a list of date:time prediction for the next few days - so a person could see when the next event might occur.
Example of prediction results:
06.01.2013 04:06pm
05.31.2013 03:29pm
05.30.2013 01:14pm
Thoughts on how to go about doing time prediction of this kind with php?

The basic answer is "no". Programming tools are not designed to do prediction. Statistical tools are designed for that purpose. You should be thinking more about R, SPSS, SAS, or some other similar tool. Some databases have rudimentary data analysis tools built-in, which is another (often inferior) option.
The standard statistical technique for time-series prediction is called ARIMA analysis (auto-regressive integrated moving average). It is unlikely that you are going to be implementing that in php/SQL. The standard statistical technique for estimating time between events is Poisson regression. It is also highly unlikely that you are going to be implementing that in php/SQL.
I observe that your data points are once per day in the evening. I might guess that this is the end of some process that runs during the day. The end time is based on the start time and the duration of the process.
What can you do? Often a reasonable prediction is "what happened yesterday". You would be surprised at how hard it is to beat this prediction for weather forecasting and for estimating the stock market. Another very reasonable method is the average of historical values.
If you know something about your process, then an average by day of the week can work well. You can also get more sophisticated, and do Monte Carlo estimates, by measuring the average and standard deviation, and then pulling a random value from a statistical distribution. However, the average value would work just as well in your case.
I would suggest that you study a bit about statistics/data mining/predictive analytics before attempting to do any "predictions". At the very least, if you really have a problem in this domain, you should be looking for the right tools to use.

As Gordon Linoff posted, the simple answer is "no", but you can write some code that will give a rough guess on what the next time will be.
I wrote a very basic example on how to do this on my site http://livinglion.com/2013/05/next-occurrence-in-datetime-sequence/

Here's a possible way that this could be done, using PHP + MySQL:
You can have a table with two fields: a DATE field and a TIME field (essentially storing the date + time portion separately). Say that the table is named "timeData" and the fields are:
eventDate: date
eventTime: time
Your primary key would be the combination of eventDate and eventTime, so that they're never repeated as a pair.
Then, you can do a query like:
SELECT eventTime, count(*) as counter FROM timeData GROUP BY eventTime ORDER BY counter DESC LIMIT 0, 10
The aforementioned query will always return the first 10 most frequent event times, ordered by frequency. You can then order these again from smallest to largest.
This way, you can return quite accurate time prediction results, which will become even more accurate as you gather data each day

Related

Convert mysql irregular time series data to regular sequence

I have a table of temperature data, updated every 5-15 mins by multiple sensors.
The data is essentially this: unique id, device(sensor id), timestamp, value(float)
The sensors does not have an accurate clock, so the readings are doomed to skew over time, so I'm unable to use things like group by hour in mysql to get a reading of the last 24h of temperature data.
My solution as a php programmer would be to make a pre-processor that reads all the un-processed readings and "join them" in a table.
There must be others than me who has this need to "downscale" x-minute/hour reads down to one per hour, to use in lets say graphing.
My problem is how do I calculate the rounded hour value from one or several readings.
For example, I have 12 readings over 2,5 hours, and I need an explicit value for each whole hour for all these readings.
Data:
Date Device Value
2016-06-27 12:15:15, TA, 23.5
2016-06-27 12:30:19, TA, 23.1
2016-06-27 12:45:35, TA, 22.9
2016-06-27 13:00:55, TA, 22.5
2016-06-27 13:05:15, TA, 22.8
2016-06-27 13:35:35, TA, 23.2
I'm not that much into statistical math, so "standard deviation" and the likes are citys in Russia for me.
Also, the devices go to sleep sometimes, and does not always transmit a temperature.
Feel free to ask me to add info to the question, as I'm not sure what you guys need to answer this.
The most important parts is this:
1. I'm using MySQL, and that's not going to change.
2. I'm hoping for a solution (or tips) in php, though tips in many other languages also would help my understanding. I'm primarily a PHP programmer though, so answers in that language would be most appreciated.
Edit: I would like to specify a few points.
Because the time data recorded from the sensors may be inaccurate, I'm relying on the SQL insert time. That way the time is controlled by one device only, the controller that's inserting the data.
For example, if I select 30 timestamp/value pairs in a 24h period, I would like to "combine" these to 24 timestamp/value pairs, using an average to combine the overflowing data.
I'm not that good to explain, but I hope this makes it clearer.
Also, would love either a clean SQL way of doing it, but also a PHP way of looping through 30 rows to produce 24 whole hour rows of data.
My goal is to have one row for every hour, with an accurate timestamp and temperature value. Mainly because most graphing libraries expect that kind of input. Especially when I have more than one series in a graph.
At some point, I may find it useful to show a graph for let's say the last six hours, with a 15 minute accuracy.
The clue is that I don't want to change the raw data, just find a way to extract/compute linear results from it.
How I would try to handle this is;
Take day start value; 01/01/2016 00:00:00 and do a 'between' 'sql' in MySQL, progressing every hour. So the first 'sql' would be like;
'select avg(temp_value) from table where date between 01/01/2016 00:00:00 and 01/01/2016 00:59:99' and progress on by the hour.
The sql isn't correct, and the entire 24hr period can be written out programmatically, but I think this will start you on your way.

Too many SQL calls on page load?

I'm constructing a website for a small collection of parents at a private daycare centre. One of the desired functions of the site is to have a calendar where you can pick what days you can be responsible for the cleaning of the locales. Now, I have made a working calendar. I found a simple script online that I modified abit to fit our purpose. Technically, it works well, but I'm starting to wonder if I really should alter the way it extracts information from the databse.
The calendar is presented monthly, and drawn as a table using a for-loop. That means that said for-loop is run 28-31 times each time the page is loaded depending on the month. To present who is responsible for cleaning each day, I have added a call to a MySQL database where each member's cleaning day is stored. The pseudo code looks like this, simplified:
Draw table month
for day=start_of_month to day=end_ofmonth
type day
select member from cleaning_schedule where picked_day=day
type member
This means that each reload of the page does at least 28 SELECT calls to the database and to me it seems both inefficient and that one might be susceptible to a DDOS-attack. Is there a more efficient way of getting the same result? There are much more complex booking calendars out there, how do they handle it?
SELECT picked_day, member FROM cleaning_schedule WHERE picked_day BETWEEN '2012-05-01' AND '2012-05-31' ORDER BY picked_day ASC
You can loop through the results of that query, each row will have a date and a person from the range you picked, in order of ascending dates.
The MySQL query cache will save your bacon.
Short version: If you repeat the same SQL query often, it will end up being served without table access as long as the underlying tables have not changed. So: The first call for a month will be ca. 35 SQL Queries, which is a lot but not too much. The second load of the same page will give back the results blazing fast from the cache.
My experience says, that this tends to be much faster than creating fancy join queries, even if that would be possible.
Not that 28 calls is a big deal but I would use a join and call in the entire month's data in one hit. You can then iterate through the MySQL Query result as if it was an array.
You can use greater and smaller in SQL. So instead of doing one select per day, you can write one select for the entire month:
SELECT day, member FROM cleaning_schedule
WHERE day >= :first_day_of_month AND day >= :last_day_of_month
ORDER BY day;
Then you need to pay attention in your program to handle multiple members per day. Although the program logic will be a bit more complex, the program will be faster: The interprocess or even network based communication is a lot slower than the additional logic.
Depending on the data structure, the following statement might be possible and more convenient:
SELECT day, group_concat(member) FROM cleaning_schedule
WHERE day >= :first_day_of_month AND day >= :last_day_of_month
GROUP BY day
ORDER BY day;
28 queries isnt a massive issue and pretty common for most commercial websites but is recommend just grabbing your monthly data by each month on one hit. Then just loop through the records day by day.

Twitter style trends with php/mysql

I am coding a social network and I need a way to list the most used trends, All statuses are stored in a content field, so what it is exactly that I need to do is match hashtag mentions such as: #trend1 #trend2 #anothertrend
And sort by them, Is there a way I can do this with MySQL? Or would I have to do this only with PHP?
Thanks in advance
The maths behind trends are somewhat complex; machine learning may be a bit over the top, but you probably need to work through some examples.
If you go with #deadtrunk's sample code, you would miss trends that have fired up in the last half hour; if you go with #eggyal's example, you miss trends that have been going strong all day, but calmed down in the last half hour.
The classic solution to this problem is to use a derivative function (http://en.wikipedia.org/wiki/Derivative); it's worth building a sample database and experimenting with this, and making your solution flexible enough to change this over time.
Whilst you want to build something simple, your users will be used to trends, and will assume it's broken if it doesn't work the way they expect.
You should probably extract the hash tags using PHP code, and then store them in your database separately from the content of the post. This way you'll be able to query them directly, rather then parsing the content every time you sort.
I think it is better to store tags in dedicated table and then perform queries on it.
So if you have a following table layout
trend | date
You'll be able to get trends using following query:
SELECT COUNT(*), trend FROM `trends` WHERE `date` = '2012-05-10' GROUP BY trend
18 test2
7 test3
Create a table that associates hashtags with statuses.
Select all status updates from some recent period - say, the last half hour - joined with the hashtag association table and group by hashtag.
The count in each group is an indication of "trend".

How to calculate percentile rank for point totals over different time spans?

On a PHP & CodeIgniter-based web site, users can earn reputation for various actions, not unlike Stack Overflow. Every time reputation is awarded, a new entry is created in a MySQL table with the user_id, action being rewarded, and value of that bunch of points (e.g. 10 reputation). At the same time, a field in a users table, reputation_total, is updated.
Since all this is sort of meaningless without a frame of reference, I want to show users their percentile rank among all users. For total reputation, that seems easy enough. Let's say my user_id is 1138. Just count the number of users in the users table with a reputation_total less than mine, count the total number of users, and divide to find the percentage of users with a lower reputation than mine. That'll be user 1138's percentile rank, right? Easy!
But I'm also displaying reputation totals over different time spans--e.g., earned in the past seven days, which involves querying the reputation table and summing all my points earned since a given date. I'd also like to show percentile rank for the different time spans--e.g., I may be 11th percentile overall, but 50th percentile this month and 97th percentile today.
It seems I would have to go through and find the reputation totals of all users for the given time span, and then see where I fall within that group, no? Is that not awfully cumbersome? What's the best way to do this?
Many thanks.
I can think of a few options off the top of my head here:
As you mentioned, total up the reputation points earned during the time range and calculate the percentile ranks based on that.
Track updates to reputation_total on a daily basis - so you have a table with user_id, date, reputation_total.
Add some new columns to the user table (reputation_total, reputation_total_today, reputation_total_last30days, etc) for each time range. You could also normalize this into a separate table (reputation_totals) to prevent you from having to add a new column for each time span you want to track.
Option #1 is the easiest, but it's probably going to get slow if you have lots of rows in your reputation transaction table - it won't scale very well, especially if you need to calculate these in real time.
Option #2 is going to require more storage over time (one row per user per day) but would probably be significantly faster than querying the transaction table directly.
Option #3 is less flexible, but would likely be the fastest option.
Both options 2 & 3 would likely require a batch process to calculate the totals on a daily basis, so that's something to consider as well.
I don't think any option is necessarily the best - they all involve different tradeoffs of speed/storage space/complexity/flexibility. What you do will ultimately depend on the requirements for your application of course.
I don't see why that would be too overly complex. Generally all you would need is to add to your WHERE clause a query that limits results like:
WHERE DatePosted between #StartOfRange and #EndOfRange

Popularity Algorithm

I'd like to populate the homepage of my user-submitted-illustrations site with the "hottest" illustrations uploaded.
Here are the measures I have available:
How many people have favourited that illustration
votes table includes date voted
When the illustration was uploaded
illustration table has date created
Number of comments (not so good as max comments total about 10 at the moment)
comments table has comment date
I have searched around, but don't want user authority to play a part, but most algorithms include that.
I also need to find out if it's better to do the calculation in the MySQL that fetches the data or if there should be a PHP/cron method every hour or so.
I only need 20 illustrations to populate the home page. I don't need any sort of paging for this data.
How do I weight age against votes? Surely a site with less submission needs less weight on date added?
Many sites that use some type of popularity ranking do so by using a standard algorithm to determine a score and then decaying eternally over time. What I've found works better for sites with less traffic is a multiplier that gives a bonus to new content/activity - it's essentially the same, but the score stops changing after a period of time of your choosing.
For instance, here's a pseudo-example of something you might want to try. Of course, you'll want to adjust how much weight you're attributing to each category based on your own experience with your site. Comments are rare, but take more effort from the user than a favorite/vote, so they probably should receive more weight.
score = (votes / 10) + comments
age = UNIX_TIMESTAMP() - UNIX_TIMESTAMP(date_created)
if(age < 86400) score = score * 1.5
This type of approach would give a bonus to new content uploaded in the past day. If you wanted to approach this in a similar way only for content that had been favorited or commented on recently, you could just add some WHERE constraints on your query that grabs the score out from the DB.
There are actually two big reasons NOT to calculate this ranking on the fly.
Requiring your DB to fetch all of that data and do a calculation on every page load just to reorder items results in an expensive query.
Probably a smaller gotcha, but if you have a relatively small amount of activity on the site, small changes in the ranking can cause content to move pretty drastically.
That leaves you with either caching the results periodically or setting up a cron job to update a new database column holding this score you're ranking by.
Obviously there is some subjectivity in this - there's no one "correct" algorithm for determining the proper balance - but I'd start out with something like votes per unit age. MySQL can do basic math so you can ask it to sort by the quotient of votes over time; however, for performance reasons, it might be a good idea to cache the result of the query. Maybe something like
SELECT images.url FROM images ORDER BY (NOW() - images.date) / COUNT((SELECT COUNT(*) FROM votes WHERE votes.image_id = images.id)) DESC LIMIT 20
but my SQL is rusty ;-)
Taking a simple average will, of course, bias in favor of new images showing up on the front page. If you want to remove that bias, you could, say, count only those votes that occurred within a certain time limit after the image being posted. For images that are more recent than that time limit, you'd have to normalize by multiplying the number of votes by the time limit then dividing by the age of the image. Or alternatively, you could give the votes a continuously varying weight, something like exp(-time(vote) + time(image)). And so on and so on... depending on how particular you are about what this algorithm will do, it could take some experimentation to figure out what formula gives the best results.
I've no useful ideas as far as the actual agorithm is concerned, but in terms of implementation, I'd suggest caching the result somewhere, with a periodic update - if the resulting computation results in an expensive query, you probably don't want to slow your response times.
Something like:
(count favorited + k) * / time since last activity
The higher k is the less weight has the number of people having it favorited.
You could also change the time to something like the time it first appeared + the time of the last activity, this would ensure that older illustrations would vanish with time.

Categories