I have never doubted the prospects and importance of big data. I also agree with its significance for the medical industry , especially now that mobile medical apps, wearable medical electronic devices, various bracelets and blood pressure monitors can bring us all. The data is collected to better serve everyone, but, after looking at the pattern of the entire industry, I want to say: Big data, it is not easy to say that it is easy to love you!
Before I talk about big data, let me talk about a book I recently read, "Signal and Noise," by US forecasting expert Nat Silver. The book says: "If the amount of information is 250 trillion per day. The speed of the bytes grows, and the useful information is definitely close to zero. Most of the information is just noise, and the noise grows much faster than the signal." From this point of view, when data information is overwhelming, We may also be getting farther and farther from the truth. The problem now is not too little data, but more data, and the valuable information we need is often drowned in a noise.
So, let me first write my conclusions first, and then discuss the correctness of this conclusion. My conclusion: Big data is not reliable, at least for now. The reasons are as follows:
First, data collection and research, with a tendency and purpose, this data is not credible.
Before we analyze the massive data, we first assume that the big data you analyze is valuable, otherwise you will not do it, but in fact, the value of this data can be collected after all the data is collected and analyzed. inferred.
Using heavily biased data can produce almost any result anyone needs. We often see that some companies say "through XX big data analysis, we draw the conclusion of YY, and our products meet the embarrassing conclusion, so how good!" If you use your own profit as the sole purpose, then his big Data analysis is especially prone to hooliganism. That is to say, if the organization that collects data can obtain economic benefits by modifying and distorting the data, then the data will be distorted and distorted. So what do we have to do with a bunch of artificially modified and distorted information?
For example, a hospital entrusts a third-party company to do a survey on the satisfaction of doctor-patient relationships. The purpose of this survey is to hang in the hospital lobby or publish it in the media for consumers to see, that is, to do Propaganda, then the survey was utilitarian at the outset. In this way, the third-party company uses various means and methods of big data research to conduct network surveys, extensively return visits, and search for comprehensive analysis from social media. The research methods are indeed true, and can be continuously tracked for a very long time to collect as much as possible. Complete data (note, popularize a concept, big data does not refer to a large amount of data, but refers to the overall study of things, comparing all the data rather than the sample analysis survey to estimate the data).
However, the so-called "patient data" collected in this way, the credibility of it is really flattering. Promoted to our industry, how many surveys have already set the tone? For a drug, a blood glucose analysis, a big data analysis report of the significance of a blood pressure monitoring, and the like.
Second, the large sample and many variables let us find the so-called correlation, and come to a conclusion without any conclusion. Ie: Finding the truth is too difficult!
One of the common mistakes of big data is that because correlation exists in many possibilities, as long as there are large samples and many variables, we may find a non-consensus correlation. It is fully in line with the strict requirements of statistical methods, but what is the relationship between the two? As long as we conduct repeated research on a pile of enough data and try different models, we will find a statistically significant correlation after thousands of times. This is another human error that is common in big data analytics—the causality is argued by relevance. It is very likely that after big data analysis, you will find that the number of birds passing by your door every day is related to the Chinese stock market.
So in the big data project, first we have to ask ourselves: Is the causal relationship assumed first and foremost?
There is often a situation in big data. There are two factors. When the data are compared with each other, it seems to have some relationship, but it is not mutually causal. Even if the two sets of data seem to be in the same rhythm, you don't know if this consistency makes sense, unless you know with certainty that the cause of one set of data will lead to another set of data. Therefore, the correlation is not necessarily a causal relationship, but it may also be due to the influence of the third factor. The so-called correlation is often verified by relevant exact values.
1. Some correlations are coincidental. Can prove to exist, but can not prove to be necessarily related.
2. The joint change between the data indicates that the two are indeed related, but it is impossible to determine which is the cause and what is the result.
Sometimes all variables have no effect on each other, but we actually find that they have significant correlation. Or, to put it another way, there is another situation where the data is real, but what is not true is the unconfirmed conclusions that are inferred from these data.
In the 2014 Baidu Alliance Summit, which was just concluded, Li Yanhong mentioned in the combination of big data and medical care : "The real big data accumulation in the future should be able to predict people's disease situation in advance, because the disease does not appear in one day, but All kinds of data must change when it is accumulated for a long time. Valuable data is not a useless information explosion, but valuable slow data, which can predict the data of personalized information." Slow data above big data, remove A lot of data from clutter interference is really valuable. So, the data is too much noise, it is too difficult to find out the truth!
Medical monitoring is a hot topic in the industry. Some monitoring technology has emerged to predict the next large-scale outbreak of epidemics such as variability by monitoring the surrounding big data and analyzing the surrounding epidemic. People no longer have to panic and worry about being infected. There is a corresponding prompt to respond to prevention in a timely manner. The medical industry closely related to this can be informed of the outbreak trend, rational allocation and deployment of medical personnel, and reminding and recommending people to prevent diseases. , reduce the probability of people getting sick... Is this seemingly beautiful and powerful? actually not.
The big data monitoring epidemic ignores the fact that big data is based on a large amount of data collected, and a series of work such as comparative research and data analysis with the existing traditional diseases in the medical field. Therefore, it is often difficult to predict unknown new diseases, such as SARS, H1N1 flu and the current Ebola outbreak. Last year, the Ministry of Health spokesperson mentioned: "China faces a serious situation where the threat of traditional epidemics persists and new epidemics continue to emerge." So we have to make an objective assessment of the disease prediction ability of big data, and advocate it on the sidelines. No value, from the system point of view, want to predict the black swan, which is itself a philosophical proposition of "mission impossible"!
Innovative jet embedded nylon technology maximizes the efficiency of sample collection from patients. Nylon adheres vertically and evenly to the surface of the swab tip, improving the efficiency of cell and liquid sample collection and release. Increased analytical sensitivity, no specimen residue, and accelerated specimen processing.
Flocked Swab,Collect swabs from specimens with flocked tips,Medical sampling swab oral nylon swab flocking,Acts in the clinical diagnosis
Jiangsu iiLO Biotechnology Co., Ltd. , https://www.sjiilobiotech.com