External Label(20220718 - 20220729)

# External Label(20220718 - 20220729) ###### tags: `external label` [TOC] pre: [External label(20220701-20220715)](https://hackmd.io/4xGI771ATD2jhe63CWqJTg) ## Target 1. y改為`Present Interest` (官網外搜尋關鍵字) `Present Interest` 單獨的字彙數量高達1000多筆，其中有許多字彙是有相同意義的，或者是某個字彙的縮寫，eg:水上遊、水上遊樂園，因此我們想先將相關的字詞進行彙整。 - 處理方法: (1)利用BERT+Kmeans進行分群(k=2...30) (2)因為Kmean對outlier很敏感，所以嘗試不同的方法來分群 spetural clustering / mini - batch k means - 選擇適合的k值: (1)Elbow method: 原本隨機分群k=30，用elbow method去看要分幾群比較好，也可以再看total within clustering variance 3. x仍維持為ncrm資料資料期間為2021/07-2021/12月，但測試資料使用2021/09 NCRM。各月NCRM與appier資料中customid有交集的客戶。約有58萬筆。 - 處理 1 : rescaling : minimax / zscore / quantile(20) ## Clustering ### Clustering methods - Euclidan Kmeans - Mini - Batch Kmeans - Spectral Clustering ### Evaluation methods - Elbow method(Distortion、inertia) : Euclidan Kmeans、Mini-batch Kmeans - silhouette coefficient : Euclidan Kmeans、Mini-batch Kmeans - Eigengap : Spectral Clustering :::spoiler **Evaluation methods for Kmeans** ### Evaluation methods for Kmeans Setting : Set min clustering = 2，max clustering =30. - **elbow method** - Criteria : If the plot looks like an arm, then the elbow on the arm is optimal k. (1) Distortion calculated as the average of the squared distances from the cluster centers of the respective clusters. Typically, the Euclidean distance metric is used. ![](https://i.imgur.com/Ikz7vkk.png =230x55) (2) Inertia sum of squared distances of samples to their closest cluster center. ![](https://i.imgur.com/xH4F6rx.png =200x90) - pros : elbow method is easy to implement and provides valuable results. - cons : 1.affected by the # of objects. 2.only calculates the euclidean distance. - **silhouette(silhouette coeffiecient)** - Criteria : The best partition is to choose K with the highest SC. - silhouette formula : It measures how close that point lies to its nearest neighbor points, across all clusters. It provides information about clustering quality which can be used to determine whether further refinement by clustering should be performed on the current clustering. The Silhouette Coefficient is calculated using the mean intra-cluster distance (`a`) and the mean nearest-cluster distance (`b`) for each sample. The Silhouette Coefficient for a sample is `(b - a) / max(a, b)`. ![](https://i.imgur.com/dMTjAPn.png =350x150) silhouette coeffiecient(=average silhouette) : ![](https://i.imgur.com/0CmiXFl.png =360x130) - pros : 1.SC is not affected by the # of objects, which is unlike Distortion. 2.takes into account variables such as variance, skewness, high-low differences - cons : The Silhouette Coefficient is generally higher for convex clusters than other concepts of clusters, such as density based clusters like those obtained through DBSCAN. - Source: https://scikit-learn.org/stable/auto_examples/cluster/plot_kmeans_silhouette_analysis.html ::: :::spoiler **Evaluation methods for Spectral Clustering** ### Evaluation methods for Spectral Clustering - why do we use Spectral Clustering : It does not require estimating an explicit model of data distribution, rather a spectral analysis of the matrix of point-to-point similarities.Useful in hard non-convex clustering problems. - Choice of number of clusters k : Most stable clustering is usually given by the value of k that maximizes the eigengap (difference between consecutive eigenvalues) - Eigengap heuristic for finding the optimal number of clusters Eigengap heuristic suggests the number of clusters k is usually given by the value of k that maximizes the eigengap (difference between consecutive eigenvalues). The larger this eigengap is, the closer the eigenvectors of the ideal case and hence the better spectral clustering works. A Tutorial on Spectral Clustering http://www.tml.cs.uni-tuebingen.de/team/luxburg/publications/Luxburg07_tutorial.pdf - Self tuning Spectral Clustering The idea behind the self tuning spectral clustering is determine the optimal number of clusters and also the similarity metric σi used in the computation of the affinity matrix. Self-Tuning Spectral Clustering(2004) : https://proceedings.neurips.cc/paper/2004/file/40173ea48d9567f1f393b20c855bb40b-Paper.pdf https://towardsdatascience.com/spectral-graph-clustering-and-optimal-number-of-clusters-estimation-32704189afbe ::: ### Clustering result - Euclidan Kmeans ![](https://i.imgur.com/LDxNeU4.png) > 從 distortion、inertia 圖中看不出來明顯的彎曲點，不同的k值的silhouette係數皆很小表示分組的狀況不太好。 - Mini - Batch Kmeans ![](https://i.imgur.com/gYNuFNU.png) > 從 distortion、inertia 圖中來看在k=20有彎曲的情形，但不同的k值的silhouette係數仍然很小表示分組的狀況不太好。 - Spectral Clustering ![](https://i.imgur.com/WsBa9c5.png) > egiengap 圖顯示 k=5,7,10,16有較高的eigangap，因此再從k=5,7,10,16中去挑選分類情形較好的k值。分得比較好的類型: 服裝、運動、旅遊、偶像、遊戲、健康、料理、寵物、行業、節日 1.mini batch kmeans 選20群的結果，從中挑選出分類結果比較穩定的8群: 服裝、運動、健康、料理、行業、偶像、旅遊、遊戲 2.spectral clustering 選10分6群: 偶像、運動、旅遊、遊戲、料理、汽車 ## Experiment ### data clean ![](https://i.imgur.com/IbfL6wA.png) > 圖為各個月 ncrm 中 feature 數值都一樣的 columns 將每個月 ncrm 中 feature 數值都一樣的剔除 (取交集), 751 -> 733 ``` useless_col = ['CREDIT_CARD_JOB_CTG_地方縣市長/地方民代','CREDIT_CARD_JOB_CTG_警政首長/局長', 'CREDIT_INS_MARK_1','BUS_LOAN_MARK_1.0','BUS_LOAN_MARK_2.0','BUS_LOAN_MARK_3.0', 'CUS_JOB_TITLE_中央(地方)參事/專門委員','CUS_JOB_TITLE_地方縣市長/地方民代', 'CUS_JOB_TITLE_警政首長/局長','DCI_LAST12M_BUY_AMT_LC','DCI_INV_NODUE_LC','DCI_LAST1_BUY_AMT_LC', 'ELOAN_JOB_TITLE_精算師','MMA_NET_CUS_MARK_2','PI_MARK_PI11','REG_SAVING_1Q_DUE_BAL_FC','RS_AVG_BAL', 'RP_AVG_BA'] ``` ### Experiment(NCRM = 9M/Kmeans clustering) :::spoiler positive rate(appier 9-11m) #### positive rate | Appier date | 服飾 | 運動| 料理| 行業| 健康| 偶像| 遊戲|旅遊| | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.476|0.757|0.544|0.756|0.764|0.372|0.708|0.439| | appier 1004 |0.44|0.725|0.53|0.83|0.714|0.344|0.692|0.466 | | appier 1022 |0.333|0.557|0.512|0.816|0.691|0.391|0.657|0.501 | | appier 1119 |0.343|0.539|0.561|0.793|0.674|0.385|0.642|0.592| ::: :::spoiler method : qt #### method : qt ##### (1)服飾 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.68|0.37|0.6|0.61|0.52|0.02|0.39|0.48|['42823', '18019', '37678', '17666']| | appier 1004 |0.68|0.38|0.49|0.58|0.54|0.01|0.25|0.44|['54596', '10267', '42679', '8644']| | appier 1022 |0.71|0.3|0.24|0.42|0.66|0.0|0.05|0.34|['75244', '1878', '38111', '953']| | appier 1119 |0.71|0.34|0.29|0.45|0.65|0.01|0.07|0.34|['73653', '2719', '38236', '1578']| ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.78|0.29|0.87|0.78|0.75|0.01|0.86|0.76|['311', '27744', '730', '87401']| | appier 1004 |0.76|0.3|0.86|0.75|0.72|0.01|0.84|0.73|['500', '31167', '1073', '83446']| | appier 1022 |0.68|0.37|0.77|0.64|0.54|0.01|0.67|0.56|['8219', '43255', '9895', '54817']| | appier 1119 |0.69|0.38|0.75|0.64|0.53|0.02|0.63|0.54|['15052', '38607', '16239', '46288']| ##### (3)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.68|0.37|0.76|0.63|0.54|0.03|0.65|0.55|['11728', '40887', '12703', '50868']| | appier 1004 |0.68|0.37|0.74|0.63|0.53|0.02|0.62|0.54|['15721', '38503', '16644', '45318']| | appier 1022 |0.68|0.37|0.72|0.63|0.52|0.03|0.57|0.52|['22997', '33515', '22437', '37237']| | appier 1119 |0.7|0.38|0.76|0.67|0.57|0.1|0.65|0.58|['18312', '32836', '17431', '47607']| ##### (4)行業 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.31|0.88|0.78|0.75|0.03|0.86|0.76|['511', '27758', '1006', '86911']| | appier 1004 |0.85|0.32|0.92|0.85|0.83|0.02|0.91|0.83|['155', '19255', '415', '96361']| | appier 1022 |0.84|0.31|0.91|0.83|0.81|0.02|0.9|0.82|['206', '21027', '507', '94446']| | appier 1119 |0.82|0.32|0.9|0.81|0.79|0.02|0.88|0.79|['297', '23773', '663', '91453']| ##### (5)健康 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.29|0.88|0.78|0.76|0.01|0.86|0.77|['283', '26937', '692', '88274']| | appier 1004 |0.75|0.3|0.85|0.74|0.71|0.01|0.83|0.72|['590', '32424', '1245', '81927']| | appier 1022 |0.73|0.31|0.84|0.72|0.69|0.01|0.81|0.69|['800', '34846', '1630', '78910']| | appier 1119 |0.72|0.32|0.83|0.71|0.67|0.01|0.8|0.68|['1190', '36522', '2163', '76311']| ##### (6)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.69|0.34|0.31|0.48|0.62|0.01|0.09|0.37|['69272', '3575', '41098', '2241']| | appier 1004 |0.7|0.3|0.24|0.43|0.65|0.0|0.05|0.35|['74043', '2063', '38967', '1113']| | appier 1022 |0.69|0.36|0.36|0.51|0.6|0.01|0.12|0.39|['65860', '4829', '42197', '3300']| | appier 1119 |0.72|0.38|0.53|0.53|0.61|0.12|0.35|0.42|['59355', '12086', '32690', '12055']| ##### (7)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.76|0.35|0.85|0.75|0.7|0.04|0.82|0.71|['1555', '32196', '2346', '80089']| | appier 1004 |0.75|0.35|0.84|0.73|0.69|0.04|0.81|0.7|['1871', '33686', '2767', '77862']| | appier 1022 |0.73|0.37|0.83|0.71|0.65|0.05|0.78|0.66|['3198', '36458', '4118', '72412']| | appier 1119 |0.71|0.35|0.81|0.69|0.63|0.04|0.77|0.65|['3635', '38031', '4700', '69820']| ##### (8)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.67|0.35|0.47|0.56|0.55|0.03|0.25|0.44|['55315', '9844', '42238', '8789']| | appier 1004 |0.68|0.37|0.55|0.6|0.53|0.03|0.34|0.48|['47347', '14341', '40384', '14114']| | appier 1022 |0.68|0.36|0.67|0.62|0.52|0.03|0.5|0.51|['31663', '26051', '30291', '28181']| | appier 1119 |0.71|0.38|0.78|0.69|0.59|0.11|0.69|0.62|['15032', '32308', '15153', '53693']| ::: :::spoiler method : raw #### method : raw ##### (1)服飾 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.68|0.37|0.59|0.61|0.52|0.02|0.38|0.48|['43070', '17772', '38146', '17198']| | appier 1004 |0.68|0.38|0.49|0.58|0.55|0.02|0.25|0.45|['54906', '9957', '42722', '8601']| | appier 1022 |0.71|0.3|0.24|0.42|0.66|0.0|0.05|0.34|['75154', '1968', '38029', '1035']| | appier 1119 |0.72|0.34|0.3|0.45|0.65|0.01|0.07|0.34|['73622', '2750', '38212', '1602']| ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.78|0.29|0.87|0.78|0.75|0.01|0.86|0.76|['277', '27778', '726', '87405']| | appier 1004 |0.76|0.31|0.86|0.75|0.72|0.02|0.84|0.73|['575', '31092', '1165', '83354']| | appier 1022 |0.68|0.39|0.77|0.64|0.54|0.01|0.67|0.56|['8590', '42884', '10429', '54283']| | appier 1119 |0.68|0.36|0.75|0.64|0.53|0.02|0.63|0.54|['14684', '38975', '16138', '46389']| ##### (3)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.67|0.37|0.75|0.63|0.54|0.03|0.66|0.55|['11546', '41069', '12484', '51087']| | appier 1004 |0.68|0.37|0.74|0.63|0.52|0.02|0.62|0.54|['15480', '38744', '16587', '45375']| | appier 1022 |0.68|0.36|0.71|0.62|0.52|0.03|0.57|0.52|['22667', '33845', '22010', '37664']| | appier 1119 |0.7|0.38|0.76|0.67|0.57|0.09|0.65|0.58|['18180', '32968', '17402', '47636']| ##### (4)行業 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.31|0.88|0.78|0.75|0.02|0.86|0.76|['425', '27844', '971', '86946']| | appier 1004 |0.85|0.33|0.92|0.85|0.83|0.02|0.91|0.83|['161', '19249', '431', '96345']| | appier 1022 |0.84|0.3|0.91|0.83|0.81|0.02|0.9|0.82|['178', '21055', '492', '94461']| | appier 1119 |0.82|0.32|0.9|0.81|0.79|0.02|0.88|0.79|['298', '23772', '707', '91409']| ##### (5)健康 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.3|0.88|0.78|0.76|0.01|0.86|0.77|['272', '26948', '690', '88276']| | appier 1004 |0.75|0.29|0.85|0.74|0.71|0.01|0.83|0.72|['531', '32483', '1181', '81991']| | appier 1022 |0.73|0.32|0.84|0.72|0.69|0.01|0.81|0.69|['870', '34776', '1618', '78922']| | appier 1119 |0.72|0.32|0.83|0.71|0.67|0.01|0.8|0.68|['1202', '36510', '2168', '76306']| ##### (6)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.7|0.34|0.32|0.48|0.62|0.01|0.1|0.37|['69268', '3579', '40998', '2341']| | appier 1004 |0.71|0.32|0.26|0.44|0.65|0.0|0.06|0.35|['73856', '2250', '38855', '1225']| | appier 1022 |0.69|0.35|0.35|0.51|0.6|0.01|0.12|0.39|['66054', '4635', '42378', '3119']| | appier 1119 |0.72|0.38|0.54|0.53|0.61|0.12|0.35|0.42|['59306', '12135', '32705', '12040']| ##### (7)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.76|0.36|0.85|0.75|0.7|0.04|0.82|0.71|['1532', '32219', '2431', '80004']| | appier 1004 |0.75|0.36|0.85|0.73|0.69|0.05|0.81|0.7|['2004', '33553', '2939', '77690']| | appier 1022 |0.72|0.35|0.82|0.71|0.65|0.05|0.78|0.66|['2989', '36667', '3921', '72609']| | appier 1119 |0.71|0.35|0.82|0.69|0.63|0.05|0.77|0.65|['3603', '38063', '4552', '69968']| ##### (8)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.68|0.36|0.47|0.57|0.55|0.03|0.25|0.45|['55418', '9741', '42163', '8864']| | appier 1004 |0.68|0.37|0.55|0.6|0.53|0.03|0.34|0.48|['47083', '14605', '40304', '14194']| | appier 1022 |0.67|0.34|0.66|0.62|0.52|0.03|0.5|0.51|['31375', '26339', '29990', '28482']| | appier 1119 |0.72|0.4|0.78|0.7|0.59|0.11|0.69|0.62|['15025', '32315', '15146', '53700']| ::: :::spoiler method : zscore #### method = zscore ##### (1)服飾 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.68|0.38|0.6|0.61 | 0.52|0.02|0.38|0.48 | ['42781', '18061', '37902', '17442']| | appier 1004 | 0.69|0.39|0.49|0.58 |0.54|0.01|0.25|0.44 | ['54670', '10193', '42732', '8591'] | | appier 1022 | 0.71|0.3|0.23|0.42| 0.66|0.0|0.05|0.34 | ['75240', '1882', '38083', '981']| | appier 1119 | 0.71|0.33|0.28|0.45 |0.65|0.01|0.07|0.34 | ['73738', '2634', '38239', '1575'] | ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.78|0.28|0.87|0.78|0.76|0.02|0.86|0.76 | ['308', '27747', '668', '87463'] || | appier 1004 | 0.76|0.31|0.86|0.75 |0.72|0.01|0.84|0.73 | ['523', '31144', '1180', '83339'] | | appier 1022 | 0.68|0.37|0.77|0.64 | 0.54|0.0|0.67|0.56 | ['8209', '43265', '10121', '54591']| | appier 1119 | 0.68|0.36|0.75|0.64 | 0.53|0.02|0.63|0.54 | ['14695', '38964', '15747', '46780']| ##### (3)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.67|0.35|0.75|0.63 | 0.54|0.03|0.66|0.55 | ['11145', '41470', '12068', '51503']| | appier 1004 |0.67|0.35|0.74|0.63 | 0.53|0.02|0.62|0.54 | ['15219', '39005', '16135', '45827'] | | appier 1022 |0.68|0.36|0.72|0.63 | 0.52|0.03|0.57|0.52 | ['22624', '33888', '22150', '37524']| | appier 1119 |0.7|0.38|0.76|0.67 | 0.57|0.09|0.65|0.58 | ['18195', '32953', '17485', '47553']| ##### (4)行業 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.31|0.88|0.78 | 0.75|0.02|0.86|0.76 | ['469', '27800', '964', '86953'] | | appier 1004 |0.85|0.32|0.92|0.85 | 0.83|0.02|0.91|0.83 | ['161', '19249', '455', '96321'] | | appier 1022 |0.84|0.3|0.91|0.83 |0.81|0.02|0.9|0.82 | ['189', '21044', '453', '94500'] | | appier 1119 |0.82|0.32|0.9|0.81 |0.79|0.02|0.88|0.79 | ['308', '23762', '686', '91430'] | ##### (5)健康 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.79|0.3|0.88|0.78| 0.76|0.02|0.86|0.77 | ['291', '26929', '655', '88311']| | appier 1004 | 0.75|0.31|0.85|0.74 |0.71|0.01|0.83|0.72 | ['582', '32432', '1259', '81913'] | | appier 1022 | 0.73|0.32|0.84|0.72 |0.69|0.01|0.81|0.69 | ['859', '34787', '1721', '78819'] | | appier 1119 | 0.72|0.32|0.83|0.71 |0.67|0.01|0.8|0.68 | ['1149', '36563', '2152', '76322']| ##### (6)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.69|0.33|0.3|0.48 | 0.62|0.01|0.09|0.37 | ['69384', '3463', '41128', '2211'] | | appier 1004 |0.71|0.32|0.26|0.44 | 0.65|0.0|0.06|0.35 | ['73880', '2226', '38839', '1241']| | appier 1022 |0.69|0.35|0.35|0.51 |0.59|0.0|0.12|0.39 | ['65948', '4741', '42333', '3164'] | | appier 1119 |0.71|0.37|0.53|0.53 |0.61|0.12|0.35|0.42 | ['59493', '11948', '32871', '11874'] | ##### (7)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.76|0.35|0.85|0.74 | 0.7|0.04|0.82|0.71 | ['1463', '32288', '2263', '80172']| | appier 1004 |0.75|0.35|0.84|0.73 | 0.69|0.05|0.81|0.7 | ['1978', '33579', '2863', '77766']| | appier 1022 |0.72|0.36|0.82|0.71 | 0.65|0.05|0.78|0.66 | ['3022', '36634', '4024', '72506']| | appier 1119 | 0.71|0.35|0.82|0.69|0.63|0.05|0.77|0.65 | ['3665', '38001', '4623', '69897'] | ##### (8)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.68|0.37|0.49|0.57 |0.55|0.03|0.26|0.44 | ['54892', '10267', '41927', '9100'] | | appier 1004 |0.67|0.36|0.55|0.6 |0.53|0.02|0.33|0.47 | ['47343', '14345', '40802', '13696'] | | appier 1022 |0.68|0.35|0.67|0.62 |0.51|0.02|0.5|0.51 | ['31541', '26173', '30507', '27965']| | appier 1119 |0.71|0.38|0.78|0.69 |0.59|0.11|0.69|0.62 | ['15095', '32245', '15149', '53697'] | ::: :::spoiler method : minimax #### method = minimax ##### (1)服飾 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.68|0.37|0.59|0.61|0.52|0.02|0.38|0.48 | ['42896', '17946', '38101', '17243'] | | appier 1004 | 0.68|0.38|0.48|0.58|0.54|0.01|0.24|0.44 | ['54785', '10078', '42952', '8371'] | | appier 1022 | 0.72|0.31|0.25|0.43|0.66|0.0|0.05|0.34 | ['75034', '2088', '37960', '1104'] | | appier 1119 | 0.72|0.34|0.3|0.45 |0.65|0.01|0.08|0.34 | ['73531', '2841', '38141', '1673']| ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.78|0.29|0.87|0.78|0.76|0.01|0.86|0.76 | ['301', '27754', '689', '87442'] | | appier 1004 | 0.76|0.3|0.86|0.75|0.72|0.01|0.84|0.73 | ['458', '31209', '1037', '83482'] | | appier 1022 | 0.67|0.36|0.76|0.63 |0.54|0.01|0.67|0.56 | ['7817', '43657', '9566', '55146'] | | appier 1119 | 0.68|0.36|0.74|0.64|0.53|0.02|0.63|0.54 | ['14406', '39253', '15729', '46798'] | ##### (3)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.67|0.36|0.75|0.63 | 0.54|0.03|0.65|0.55 | ['11656', '40959', '12760', '50811'] | | appier 1004 |0.67|0.36|0.74|0.63 | 0.53|0.03|0.63|0.54 | ['15384', '38840', '16076', '45886']| | appier 1022 |0.69|0.38|0.72|0.63 | 0.52|0.03|0.57|0.52 | ['23139', '33373', '22602', '37072'] | | appier 1119 |0.69|0.36|0.75|0.66 | 0.57|0.1|0.66|0.58 | ['18056', '33092', '17182', '47856']| ##### (4)行業 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.31|0.88|0.78 |0.75|0.02|0.86|0.76 | ['472', '27797', '979', '86938'] | | appier 1004 | 0.85|0.32|0.92|0.85| 0.83|0.02|0.91|0.83 | ['149', '19261', '408', '96368']| | appier 1022 | 0.84|0.31|0.91|0.83 | 0.81|0.01|0.9|0.82 | ['176', '21057', '516', '94437'] | | appier 1119 | 0.82|0.31|0.9|0.81|0.79|0.02|0.88|0.79 | ['298', '23772', '705', '91411'] | ##### (5)健康 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.79|0.3|0.88|0.79 |0.76|0.01|0.86|0.77 | ['275', '26945', '707', '88259'] | | appier 1004 |0.75|0.3|0.85|0.74 |0.71|0.01|0.83|0.72 | ['544', '32470', '1198', '81974'] | | appier 1022 | 0.73|0.32|0.84|0.72 |0.69|0.01|0.81|0.69 | ['887', '34759', '1716', '78824']| | appier 1119 |0.72|0.32|0.83|0.71 | 0.67|0.01|0.8|0.68 | ['1214', '36498', '2157', '76317']| ##### (6)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.69|0.33|0.31|0.48 |0.62|0.01|0.09|0.37 | ['69305', '3542', '41116', '2223'] | | appier 1004 |0.7|0.31|0.25|0.43 | 0.65|0.0|0.05|0.35 | ['73871', '2235', '38885', '1195']| | appier 1022 |0.69|0.35|0.35|0.51 | 0.6|0.01|0.12|0.39 | ['66016', '4673', '42312', '3185']| | appier 1119 |0.72|0.38|0.54|0.53 | 0.61|0.12|0.35|0.42 | ['59200', '12241', '32557', '12188']| ##### (7)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.76|0.35|0.85|0.74 | 0.7|0.04|0.82|0.71 | ['1481', '32270', '2302', '80133']| | appier 1004 |0.75|0.36|0.84|0.73 | 0.69|0.05|0.81|0.7 | ['2033', '33524', '2872', '77757'] | | appier 1022 |0.72|0.35|0.82|0.7 |0.65|0.05|0.78|0.66 | ['3081', '36575', '3960', '72570'] | | appier 1119 |0.71|0.36|0.82|0.7 |0.63|0.05|0.77|0.65 | ['3655', '38011', '4664', '69856'] | ##### (8)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.68|0.37|0.48|0.57| 0.55|0.03|0.26|0.45 | ['55152', '10007', '42044', '8983'] | | appier 1004 |0.68|0.37|0.55|0.6 | 0.53|0.03|0.34|0.48 | ['47437', '14251', '40639', '13859'] | | appier 1022 |0.68|0.36|0.67|0.62 | 0.51|0.03|0.5|0.51 | ['31167', '26547', '29808', '28664'] | | appier 1119 |0.72|0.4|0.79|0.69 |0.59|0.11|0.69|0.62 | ['14717', '32623', '14971', '53875'] | ::: ### Experiment(NCRM = 9M/Spectral Clustering) 由於後來做出 spectral clustering 的方法, 我們覺得它分 10 類中的其中的 6 類結果不錯, 因此後面的 label 都會是用這六類的結果 ``` ["偶像","運動","旅遊","遊戲","料理","汽車"] ``` ``` {'冰上曲棍', '台灣服', '夏威夷旅', '拼字', '男性話題', '白色', '白色情', '虛擬貨幣', '貓用', '阿聯酋旅'} ``` #### positive rate | Appier date |偶像|運動|旅遊|遊戲|料理|汽車| | ----------- | - | -- | - | - | - | -- | | appier 0922 |0.378|0.722|0.334|0.735|0.37|0.109 | appier 1004 |0.35|0.681|0.371|0.717|0.357|0.119 | appier 1022 |0.401|0.411|0.408|0.683|0.403|0.134 | appier 1119 |0.39|0.382|0.518|0.716|0.46|0.127 :::spoiler method : zscore #### method = zscore ##### (1)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.69|0.33|0.3|0.48 |0.62|0.01|0.09|0.37 | ['69384', '3463', '41128', '2211'] | | appier 1004 |0.71|0.32|0.26|0.44 |0.65|0.0|0.06|0.35 | ['73880', '2226', '38839', '1241'] | | appier 1022 |0.69|0.35|0.35|0.51 |0.59|0.0|0.12|0.39 | ['65948', '4741', '42333', '3164'] | | appier 1119 |0.71|0.37|0.53|0.53 |0.61|0.12|0.35|0.42 | ['59493', '11948', '32871', '11874'] | ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.75|0.31|0.85|0.74 | 0.71|0.01|0.83|0.72 | ['619', '32576', '1240', '81751']| | appier 1004 |0.72|0.33|0.83|0.71 | 0.66|0.01|0.8|0.67 | ['1145', '36792', '2180', '76069'] | | appier 1022 |0.69|0.37|0.39|0.53 |0.58|0.01|0.15|0.41 | ['63174', '5990', '42618', '4404']| | appier 1119 |0.7|0.36|0.34|0.49 | 0.61|0.02|0.12|0.38 | ['68346', '4325', '40556', '2959']| ##### (3)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.72|0.31|0.25|0.42 | 0.66|0.01|0.05|0.33 | ['76107', '2031', '36991', '1057'] | | appier 1004 | 0.7|0.34|0.31|0.47|0.62|0.01|0.09|0.37 | ['70371', '3416', '40229', '2170'] | | appier 1022 | 0.68|0.34|0.35|0.51|0.59|0.02|0.14|0.4 | ['64649', '5067', '42678', '3792'] | | appier 1119 | 0.71|0.42|0.71|0.66|0.61|0.21|0.61|0.58 | ['35167', '21171', '24703', '35145'] | ##### (4)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.77|0.34|0.86|0.76 |0.72|0.04|0.83|0.73 | ['1250', '30835', '1985', '82116'] | | appier 1004 | 0.76|0.36|0.85|0.74 |0.7|0.04|0.82|0.71 | ['1670', '32590', '2560', '79366'] | | appier 1022 |0.73|0.37|0.83|0.72 |0.66|0.05|0.79|0.67 | ['2646', '35773', '3657', '74110'] | | appier 1119 |0.75|0.34|0.85|0.73 | 0.69|0.03|0.81|0.7 | ['1511', '33761', '2382', '78532']| ##### (5)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.7|0.34|0.32|0.48 |0.63|0.02|0.1|0.37 | ['70398', '3465', '40031', '2292'] | | appier 1004 | 0.71|0.34|0.3|0.46| 0.64|0.0|0.07|0.35 | ['72517', '2980', '39007', '1682']| | appier 1022 | 0.69|0.36|0.38|0.52 | 0.59|0.02|0.14|0.4 | ['65012', '5344', '41921', '3909']| | appier 1119 | 0.71|0.4|0.66|0.61| 0.6|0.18|0.53|0.51 | ['43287', '19957', '26712', '26230']| ##### (6)汽車 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.9|0.29|0.17|0.19|0.89|0.0|0.0|0.11 | ['103747', '61', '12369', '9'] | | appier 1004 | 0.89|0.28|0.16|0.19| 0.88|0.0|0.0|0.12 | ['102547', '90', '13535', '14'] | | appier 1022 | 0.88|0.28|0.16|0.21| 0.87|0.0|0.0|0.13 | ['100877', '108', '15179', '22'] | | appier 1119 | 0.89|0.29|0.17|0.2| 0.88|0.01|0.0|0.12 | ['101907', '92', '14161', '26']| ::: :::spoiler method : minimax #### method = minimax ##### (1)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.69|0.33|0.31|0.48 | 0.62|0.01|0.09|0.37 | ['69305', '3542', '41116', '2223']| | appier 1004 |0.7|0.31|0.25|0.43 |0.65|0.0|0.05|0.35 | ['73871', '2235', '38885', '1195'] | | appier 1022 |0.69|0.35|0.35|0.51|0.6|0.01|0.12|0.39 | ['66016', '4673', '42312', '3185']| | appier 1119 | 0.72|0.38|0.54|0.53| 0.61|0.12|0.35|0.42 | ['59200', '12241', '32557', '12188']| ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.75|0.31|0.85|0.74 |0.71|0.02|0.83|0.72 | ['659', '32536', '1282', '81709'] | | appier 1004 |0.72|0.32|0.83|0.7 | 0.66|0.01|0.8|0.67 | ['1157', '36780', '2180', '76069']| | appier 1022 |0.69|0.37|0.4|0.53 |0.58|0.01|0.16|0.41 | ['63033', '6131', '42438', '4584'] | | appier 1119 |0.7|0.36|0.35|0.5 |0.61|0.02|0.12|0.38 | ['68362', '4309', '40588', '2927'] | ##### (3)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.72|0.32|0.27|0.43 |0.66|0.01|0.06|0.33 | ['75959', '2179', '36888', '1160'] | | appier 1004 |0.7|0.33|0.31|0.47 | 0.62|0.01|0.09|0.37 | ['70444', '3343', '40282', '2117']| | appier 1022 |0.69|0.37|0.39|0.53 | 0.59|0.01|0.15|0.4 | ['64050', '5666', '42389', '4081']| | appier 1119 |0.71|0.42|0.71|0.66 | 0.61|0.21|0.61|0.58 | ['34963', '21375', '24512', '35336']| ##### (4)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.77|0.35|0.86|0.76 |0.72|0.04|0.83|0.73 | ['1282', '30803', '1989', '82112'] | | appier 1004 |0.75|0.34|0.85|0.74| 0.7|0.05|0.82|0.71 | ['1593', '32667', '2340', '79586']| | appier 1022 |0.73|0.36|0.83|0.72|0.66|0.05|0.79|0.67 | ['2669', '35750', '3633', '74134'] | | appier 1119 |0.74|0.33|0.84|0.73 |0.69|0.04|0.81|0.7 | ['1545', '33727', '2363', '78551'] | ##### (5)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.71|0.36|0.34|0.48 | 0.62|0.01|0.1|0.37 | ['70120', '3743', '39935', '2388']| | appier 1004 |0.71|0.33|0.28|0.45 | 0.64|0.01|0.08|0.35 | ['72636', '2861', '38948', '1741']| | appier 1022 |0.69|0.36|0.38|0.52 | 0.59|0.02|0.14|0.4 | ['65004', '5352', '41909', '3921']| | appier 1119 |0.71|0.41|0.66|0.61 |0.6|0.19|0.53|0.51 | ['43440', '19804', '26728', '26214'] | ##### (6)汽車 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.9|0.28|0.16|0.19 | 0.89|0.0|0.0|0.11 | ['103743', '65', '12366', '12']| | appier 1004 | 0.89|0.27|0.16|0.19 |0.88|0.0|0.0|0.12 | ['102565', '72', '13534', '15'] | | appier 1022 | 0.88|0.27|0.16|0.2|0.87|0.01|0.0|0.13 | ['100864', '121', '15174', '27'] | | appier 1119 |0.89|0.27|0.15|0.2 |0.88|-0.0|0.0|0.12 | ['101889', '110', '14175', '12'] | ::: :::spoiler method : qt #### method = qt ##### (1)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.69|0.34|0.31|0.48| 0.62|0.01|0.09|0.37 | ['69272', '3575', '41098', '2241']| | appier 1004 | 0.7|0.3|0.24|0.43|0.65|0.0|0.05|0.35 | ['74043', '2063', '38967', '1113'] | | appier 1022 | 0.69|0.36|0.36|0.51|0.6|0.01|0.12|0.39 | ['65860', '4829', '42197', '3300'] | | appier 1119 | 0.72|0.38|0.53|0.53|0.61|0.12|0.35|0.42 | ['59355', '12086', '32690', '12055'] | ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.75|0.32|0.85|0.74 |0.71|0.0|0.83|0.71 | ['615', '32580', '1420', '81571'] | | appier 1004 |0.72|0.33|0.83|0.71 |0.66|0.01|0.8|0.67 | ['1216', '36721', '2222', '76027'] | | appier 1022 |0.69|0.36|0.38|0.53 |0.58|0.01|0.15|0.41 | ['63327', '5837', '42699', '4323'] | | appier 1119 |0.7|0.36|0.34|0.49|0.61|0.02|0.12|0.38 | ['68514', '4157', '40601', '2914']| ##### (3)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.72|0.31|0.25|0.42 | 0.66|0.01|0.05|0.33 | ['76015', '2123', '36916', '1132']| | appier 1004 |0.7|0.34|0.31|0.47 |0.63|0.01|0.09|0.37 | ['70426', '3361', '40191', '2208']| | appier 1022 |0.69|0.36|0.37|0.52|0.59|0.01|0.14|0.4 | ['64280', '5436', '42553', '3917']| | appier 1119 |0.71|0.42|0.71|0.66|0.61|0.21|0.61|0.58 | ['35295', '21043', '24662', '35186']| ##### (4)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.77|0.35|0.86|0.76 |0.72|0.04|0.83|0.73 | ['1255', '30830', '1984', '82117'] | | appier 1004 | 0.75|0.35|0.85|0.74 |0.7|0.05|0.82|0.71 | ['1714', '32546', '2510', '79416'] | | appier 1022 | 0.73|0.36|0.83|0.72 |0.66|0.04|0.79|0.67 | ['2566', '35853', '3631', '74136'] | | appier 1119 | 0.74|0.33|0.84|0.73 |0.69|0.03|0.81|0.7 | ['1471', '33801', '2337', '78577']| ##### (5)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.71|0.36|0.34|0.49|0.62|0.01|0.1|0.37 | ['70084', '3779', '39906', '2417'] | | appier 1004 | 0.71|0.33|0.29|0.45|0.64|0.01|0.07|0.35 | ['72626', '2871', '39034', '1655'] | | appier 1022 | 0.69|0.35|0.37|0.51|0.59|0.01|0.14|0.4 | ['65048', '5308', '42027', '3803'] | | appier 1119 | 0.7|0.4|0.66|0.61 |0.6|0.19|0.53|0.51 | ['43403', '19841', '26600', '26342']| ##### (6)汽車 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.9|0.29|0.17|0.19|0.89|0.0|0.0|0.11 | ['103724', '84', '12365', '13'] | | appier 1004 | 0.89|0.28|0.16|0.19|0.88|0.01|0.0|0.12 | ['102547', '90', '13527', '22'] | | appier 1022 | 0.88|0.28|0.16|0.21|0.87|0.01|0.0|0.13 | ['100882', '103', '15172', '29'] | | appier 1119 | 0.89|0.29|0.17|0.2|0.88|0.0|0.0|0.12 | ['101888', '111', '14167', '20']| ::: :::spoiler method : raw #### method = raw ##### (1)偶像 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.7|0.34|0.32|0.48| 0.62|0.01|0.1|0.37 | ['69268', '3579', '40998', '2341']| | appier 1004 | 0.71|0.32|0.26|0.44| 0.65|0.0|0.06|0.35 | ['73856', '2250', '38855', '1225']| | appier 1022 | 0.69|0.35|0.35|0.51 | 0.6|0.01|0.12|0.39 | ['66054', '4635', '42378', '3119']| | appier 1119 | 0.72|0.38|0.54|0.53|0.61|0.12|0.35|0.42 | ['59306', '12135', '32705', '12040'] | ##### (2)運動 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.75|0.3|0.85|0.74| 0.71|0.01|0.83|0.72 | ['622', '32573', '1245', '81746']| | appier 1004 | 0.72|0.32|0.83|0.71| 0.66|0.01|0.8|0.67 | ['1143', '36794', '2157', '76092']| | appier 1022 | 0.69|0.37|0.39|0.53 | 0.58|0.01|0.16|0.41 | ['63097', '6067', '42506', '4516']| | appier 1119 | 0.7|0.36|0.36|0.5 | 0.61|0.02|0.12|0.38 | ['68302', '4369', '40459', '3056']| ##### (3)旅遊 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.72|0.33|0.27|0.43 |0.66|0.01|0.06|0.33 | ['75920', '2218', '36890', '1158'] | | appier 1004 |0.7|0.33|0.3|0.47 | 0.63|0.01|0.09|0.37 | ['70529', '3258', '40307', '2092']| | appier 1022 |0.68|0.35|0.36|0.51 | 0.59|0.02|0.14|0.4 | ['64485', '5231', '42582', '3888']| | appier 1119 |0.71|0.42|0.71|0.67 | 0.6|0.21|0.6|0.58 | ['35151', '21187', '24756', '35092'] | ##### (4)遊戲 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.77|0.34|0.86|0.76|0.72|0.04|0.83|0.73 | ['1282', '30803', '2019', '82082'] | | appier 1004 | 0.75|0.35|0.85|0.74 |0.7|0.05|0.82|0.71 | ['1715', '32545', '2536', '79390'] | | appier 1022 | 0.73|0.35|0.83|0.71| 0.66|0.05|0.79|0.67 | ['2555', '35864', '3463', '74304'] | | appier 1119 | 0.74|0.34|0.84|0.73| 0.69|0.03|0.81|0.7 | ['1443', '33829', '2273', '78641']| ##### (5)料理 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 | 0.71|0.36|0.34|0.48| 0.62|0.01|0.1|0.37 | ['70010', '3853', '39833', '2490']| | appier 1004 | 0.71|0.34|0.3|0.46| 0.64|0.01|0.08|0.35 | ['72581', '2916', '38921', '1768']| | appier 1022 | 0.7|0.37|0.39|0.52 | 0.59|0.02|0.15|0.4 | ['64881', '5475', '41810', '4020']| | appier 1119 | 0.7|0.4|0.66|0.61 | 0.6|0.19|0.53|0.51 | ['43400', '19844', '26643', '26299']| ##### (6)汽車 | Appier date | tr_acc | tr_mcc | tr_f1 | tr_ap | te_acc | te_mcc | te_f1 | te_ap | te_cm | | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | -------- | | appier 0922 |0.9|0.29|0.17|0.19 | 0.89|0.0|0.0|0.11 | ['103756', '52', '12371', '7']| | appier 1004 | 0.9|0.29|0.17|0.2| 0.88|-0.0|0.0|0.12 | ['102551', '86', '13541', '8']| | appier 1022 | 0.88|0.28|0.16|0.21|0.87|0.0|0.0|0.13 | ['100864', '121', '15177', '24']| | appier 1119 | 0.89|0.28|0.16|0.2 |0.88|0.0|0.0|0.12 | ['101903', '96', '14171', '16']| ::: <br> :::success Conclusion: appier 1119 在 `料理`、`偶像`、`旅遊` 的 mcc 都明顯比其他月份的 appier 好, 因此想多試其他月份的 appier 與 ncrm, 看看是因為 ncrm 與 appier 資料不重疊所以較好, 還是 appier 1119 資料比較特別。在資料處理的方面，我們使用四種方式(zscore、minimax、qt、raw)，但每種方法的結果相差不大，可能是因為訓練不起來。 -> 新增其他月份的NCRM(2021/07-2021/12)、appier(20211105、20220103、20220314) -> 重新取用戶的交集 ::: next: [External Label(20220801 - 20220805)](https://hackmd.io/uARNJ8EmS0epMR45SO20ig)

Syntax	Example	Reference
# Header	Header	基本排版
- Unordered List	Unordered List
1. Ordered List	Ordered List
- [ ] Todo List	Todo List
> Blockquote	Blockquote
Bold font	Bold font
Italics font	Italics font
~~Strikethrough~~	~~Strikethrough~~
19^th^	19^th
H~2~O	H₂O
++Inserted text++	Inserted text
==Marked text==	Marked text
[link text](https:// "title")	Link
![image alt](https:// "title")	Image
`Code`	`Code`	在筆記中貼入程式碼
```javascript var i = 0; ```	`var i = 0;`
:smile:		Emoji list
{%youtube youtube_id %}	Externals
$L^aT_eX$	L^aT_eX
:::info This is a alert area. :::	This is a alert area.