みなさんこんにちは。3日フォースアーゼアです。どうですか?楽しみ?素晴らしい?今日は私のセッションについてお願いいたします。今日はコロナウイルスのために楽しみましょう。私のプレゼタションを楽しみましょう。では、私の話を始めましょう。まず、私の話を話します。みなさんこんにちは。私はケンジカーノーベー。私は日本のサイエンス・ソフトウェア・エンジニアです。最近、私はデッド・サイエンス・パイソン・ジャンゴ・ビュージュエース・ファイバスを勉強しました。この時、私はザーストリーを作っています。次に、私のコンポリーについてお話しします。日本のサイエンス・ラボロイトリーのコンポリーは何ですか?私はJSLのコンポリーです。JSLはNaganoプリフェクトリーです。日本では、私のコンポリーは free community space called GigLabです。Nagano is a very nice place surrounded by nature.もし、Naganoに来ても、私のコンポリーでGigLabに来てもらえます。では、私の紹介は終わります。私はラインボートについてお話しします。LINEはLINEコープレーションでのメッセージアプリケーションです。ほとんど日本人はラインを使っています。私は何を使っていますか?次に、私のラインボートの表現を説明します。ラインボートは、私のバークを取り出す場合、バークセフトを取り出す場合、しっかり、バイクを取り出す場合。これは、バークセフトのより 機能が使うことができます。次に、私のビデオ用についてお話しします。まず、このバークは友人でお見せします。そして、あなたが彼らと移住してもらえます。あなたが彼らと移住してもらえます。とても同じことができます。キーボードを使わないといけません。次に、サーフティスクールを知りたい場合は、バイクにカクレートボタンを選択します。その後、ロックスデータやロケーションデートを選択します。そして、サーフティスクールをカクレートします。はい。次に、コンフィギュラーションのシステムのダイアグラムを説明します。このシステムは私のシステムです。Ideployed Jungle on Hex.Database uses post-war SQL.LineMessage API is a useful message in line.This API is released by Line Corporation.It enables two ways of communication between system and line users.It has pushMessage and responseMessage.PushMessage is that system sends message to usersand replyMessage is that user sends message to systemand system repliesMessage to user.MySystem uses replyMessage.I'll return to explanation of the configuration diagram.OK.First, JungleView.py gets user data with using that message API.And then the user data is stored in Database by a model.And then when ordered to calculate safety school,Views.py calculates it from user dataand open data with store people for predicting theft bike.And then sends the calculated safety schoolto user by messaging API.So, but what is safety school?Next, I explain detail of safety school.The safety school is the degree of similarity store people and the user.High safety school indicates that the user datais not similar to the future of store people.It means safe.On the other hand,low safety school indicates thatuser data is similar to the future of store people.It means danger.This time to calculate safety school,I use Mahalo-based distance.It can be calculatedfrom the mean value of data setand variance of data set.It is number that indicateshow far a point is from the data set.So,repress a point with a userand replace data set with open datawith store people.Therefore,indicating how far a useris from open data is indicatinghow similar a user and store people.From this result,I use Mahalo-based distanceto calculate safety school.Let's take a closer look at the detailof the open data to usecalculation of safety school.This is open data with store peoplepublished on the website of Nagant Prefecture.There are 1,496 casesof bag theft in 2018.The open data has several columns.For example,my spell code and location,date,oct time,and age occupationof store peopleand log status.This time,I use calumsto calculate safety school,oct date,and oct time,and store people age,oct occupation,and log status of bag.So then,I use the ocalculation.Next doing is a bit annoyingbecause I must digitize them.This column is not only numerical data,have character data.Oh my god,there's only Japanese wordin occupation columns.I forgot that I am Japanese.It cannot be usedfor calculation as is.So let's digitize for calculating safety school.I use Jupyter Notebookto digitize this open data.At first,let's open data.calums not used this timein the way,so there is them.It is difficult programif comes names remains Japanese.So replace them with English.It became a little easier to see.Next,digitize is log column.Check the category value in the column.This means locked.On the other hand,this means not locked.So replace locked with 1and not locked with 0.That is this.Age and occupation countsin the same way,digitize.This goes in concrete term.Check the category,replace categoryand digitize.So about orca date time explain.Only mouse data are extractedfrom orca date and then digitize.And orca time only converts to a number.About location data explain.I added latitude and longitude columnby using CSV according servicethat converts other data columnsto latitude and longitude.Here is digitized result.No character,no Japanese words.It's clear than before.So my system using that open datahas been digitized in advance.Next,I explain how to codefor calculating safety score.As I said earlier,I use my hand with distance.At first,prepare mean vector.The mean vector is calculatedincluding the user data.So out of the user data,at the end of the open data.By the way,this user datais an arbitrary value.So my system,that is a valueobtained from line or start in the debut.And next,great constantsfor columns and loads.Then calculate the average valuefor each column using for loop.Mean vector has been calculated.Next,I calculate variance for variance.So to calculate that value,use a vector of user data minus mean vector.Then assign that valueof variable diff.Untransport that vector.If you use namifyswapoxys,you can calculatetransport vector in one line.Then multiply thosevalues and divide by number of loads.This is point result.This is print result.Now that variance for variance has been calculated.Next,calculate mohalanobis.First,calculate the inverse matrixof variance for variance.Then,multiply that inverse matrixand the vector of user data minus mean vector.Assign this value to variable var.Next,multiplythat variable andtransport that vector.Finally,the mohalanobis distance is calculatedby taking the square root of this value.The mohalanobis value this timeis 8.79.This time,the user data setabitually is that okacham is 20pm.Occupation is 4.It means working adult.Locked status is 1.It means locked.etc.As an experiment,I set the age only to 30.Calculate the safety score.The score increases a little.Set the age only to 60.The score increases considerably.Since this result,I create a graphto observe the changein the safety score by mohalanobis.This graph showsthe safety score by mohalanobisfor 1,490 storey peoplein open data.Looking at the features of people with low safety scores,I found thatthere are a lot of teens and 20sthen the time are a lot of 7am8am and that day ismany friday.On the other hand,looking at the features of people with high safety scores,I found thatthere are a lot of 60s and 70s.The relation betweenexcept and this feature is really light.So,I checked the open data I used this time.But it is difficult to find the featuresif they are in the table as is.So,I visualized them so thatfeature can be easily searched.Yes,I created some graphs.This time,I will explain featuresand trends found from these two graphs.First,this graphshows number of storey peopleby age occupation.From this graph,I found thatnumber of teens and 20s were storey bikeis over two thirds of the whole.On the other hand,number of 60s and 70swere storey bike is very small.Next,I explained the graph in the lower right.It is a graph called Hitmap.This graph shows number of storey peopleby time and day.The vertical axis is timeand the horizontal axis is day.Deep colored area means a high numberof safeties.The point surrounded by red circle is deep colored.So,check the time and day.Time in the point is 7amand 8amand 18pm.They are commiting hours in Japanand 18pm is go home hours.Next,please look at the day.The day in the point is mainly Friday.As a result of looking at the graph,I found thatstorey people have several features.Most of the storey people are teens and 20s.60s and 70s bike isdifficult to be storey.And commiting hours,go home hoursmany friday.The features that safety school by Mahanabes has judgeddanger isalmost matched features that open datadiscollect as danger.So that I use the safety school by Mahanabesfor predicting back theft.finally,conclusion.There are several features of people store bike.My linebot use open data with storey peopleand can predict back theft.In open data with store people,many younger are store bike.But I guess thatyoung people buy more high performance bikeso that to improve the accuracyof safety school,the open datashould have data about bike.That brings me to the end of all.Oh,I forgotto tell you.Code of my linebot is of course openbecause today isForce Azure.Anyone can view my code.Anyone can use it.Please try using it.If you have any questions,pleasefeel free to contact me onSNS.at kenji7175That brings me to the endof my presentation.I hope you have enjoyed.Thank you for your attention.Finish